[ { "hf_id": "ByteDance-Seed/Seed-OSS-36B-Instruct", "hf_org": "ByteDance-Seed", "hf_repo": "Seed-OSS-36B-Instruct", "title": "Seed-OSS-36B", "provider": "Seed (ByteDance)", "description": "ByteDance Seed-OSS 36B dense model with unique 'thinking budget' control and 512K context support", "date_updated": "2026-04-17", "difficulty": "intermediate", "min_vllm_version": "0.11.0", "architecture": "dense", "parameter_count": "36B", "active_parameters": "36B", "context_length": 524288, "tasks": [ "text" ], "variants": { "default": { "precision": "bf16", "vram_minimum_gb": 86 } }, "compatible_strategies": [ "single_node_tp", "multi_node_tp", "multi_node_tp_pp" ], "features": [ "tool_calling" ], "opt_in_features": [], "url": "/ByteDance-Seed/Seed-OSS-36B-Instruct", "json": "/ByteDance-Seed/Seed-OSS-36B-Instruct.json" }, { "hf_id": "Google/gemma-4-31B-it", "hf_org": "Google", "hf_repo": "gemma-4-31B-it", "title": "Gemma 4 31B IT", "provider": "Google", "description": "Google's unified multimodal Gemma 4 dense model (31B) with native text, image, and audio, plus thinking mode and tool-use protocol.", "date_updated": "2026-04-17", "difficulty": "intermediate", "min_vllm_version": "0.12.0", "architecture": "dense", "parameter_count": "31B", "active_parameters": "31B", "context_length": 262144, "tasks": [ "multimodal", "text" ], "performance_headline": "Unified multimodal model with structured thinking, function calling, dynamic vision resolution", "variants": { "default": { "precision": "bf16", "vram_minimum_gb": 210 }, "nvfp4": { "precision": "nvfp4", "vram_minimum_gb": 19 } }, "compatible_strategies": [ "single_node_tp", "multi_node_tp", "multi_node_tp_pp" ], "features": [ "tool_calling", "reasoning" ], "opt_in_features": [], "url": "/Google/gemma-4-31B-it", "json": "/Google/gemma-4-31B-it.json" }, { "hf_id": "Google/translategemma-27b-it", "hf_org": "Google", "hf_repo": "translategemma-27b-it", "title": "TranslateGemma 27B IT", "provider": "Google", "description": "Lightweight open translation model from Google (based on Gemma 3) supporting 55 languages. Served via the vLLM-optimized Infomaniak-AI checkpoint.", "date_updated": "2026-04-17", "difficulty": "beginner", "min_vllm_version": "0.14.1", "architecture": "dense", "parameter_count": "27B", "active_parameters": "27B", "context_length": 131072, "tasks": [ "text" ], "performance_headline": "Deployable on laptops/desktops and cloud GPUs; vLLM-optimized checkpoint removes custom JSON inputs", "variants": { "default": { "precision": "bf16", "vram_minimum_gb": 65 }, "vllm_optimized": { "precision": "bf16", "vram_minimum_gb": 65 }, "small_4b": { "precision": "bf16", "vram_minimum_gb": 10 } }, "compatible_strategies": [ "single_node_tp", "multi_node_tp", "multi_node_tp_pp" ], "features": [], "opt_in_features": [], "url": "/Google/translategemma-27b-it", "json": "/Google/translategemma-27b-it.json" }, { "hf_id": "MiniMaxAI/MiniMax-M2.1", "hf_org": "MiniMaxAI", "hf_repo": "MiniMax-M2.1", "title": "MiniMax-M2.1", "provider": "MiniMax", "description": "MiniMax M2.1 MoE language model (230B total / 10B active) for coding, agent toolchains, and long-context reasoning — native FP8 checkpoint", "date_updated": "2026-04-17", "difficulty": "intermediate", "min_vllm_version": "0.11.0", "architecture": "moe", "parameter_count": "230B", "active_parameters": "10B", "context_length": 196608, "tasks": [ "text" ], "performance_headline": "Updated M2 series MoE with strong SWE-Bench and Terminal-Bench performance, 196K context", "variants": { "default": { "precision": "fp8", "vram_minimum_gb": 276 } }, "compatible_strategies": [ "single_node_tp", "single_node_tep", "single_node_dep", "multi_node_tp", "multi_node_tp_pp", "multi_node_tep", "multi_node_dep" ], "features": [ "tool_calling", "reasoning" ], "opt_in_features": [], "url": "/MiniMaxAI/MiniMax-M2.1", "json": "/MiniMaxAI/MiniMax-M2.1.json" }, { "hf_id": "MiniMaxAI/MiniMax-M2.5", "hf_org": "MiniMaxAI", "hf_repo": "MiniMax-M2.5", "title": "MiniMax-M2.5", "provider": "MiniMax", "description": "MiniMax M2.5 MoE language model (230B total / 10B active) for coding, agent toolchains, and long-context reasoning — native FP8 checkpoint", "date_updated": "2026-04-17", "difficulty": "intermediate", "min_vllm_version": "0.19.0", "architecture": "moe", "parameter_count": "230B", "active_parameters": "10B", "context_length": 196608, "tasks": [ "text" ], "performance_headline": "Refreshed M2 series MoE with strong SWE-Bench and Terminal-Bench performance, 196K context", "variants": { "default": { "precision": "fp8", "vram_minimum_gb": 276 }, "nvfp4": { "precision": "nvfp4", "vram_minimum_gb": 138 } }, "compatible_strategies": [ "single_node_tp", "single_node_tep", "single_node_dep", "multi_node_tp", "multi_node_tp_pp", "multi_node_tep", "multi_node_dep" ], "features": [ "tool_calling", "reasoning" ], "opt_in_features": [], "url": "/MiniMaxAI/MiniMax-M2.5", "json": "/MiniMaxAI/MiniMax-M2.5.json" }, { "hf_id": "MiniMaxAI/MiniMax-M2.7", "hf_org": "MiniMaxAI", "hf_repo": "MiniMax-M2.7", "title": "MiniMax-M2.7", "provider": "MiniMax", "description": "MiniMax M2.7 MoE language model (230B total / 10B active) — latest M2 release for coding, agent toolchains, and long-context reasoning with native FP8", "date_updated": "2026-04-17", "difficulty": "intermediate", "min_vllm_version": "0.20.0", "architecture": "moe", "parameter_count": "230B", "active_parameters": "10B", "context_length": 196608, "tasks": [ "text" ], "performance_headline": "Latest M2 series release; verified accuracy on AIME25, GPQA-D, GSM8K; 196K context", "variants": { "default": { "precision": "fp8", "vram_minimum_gb": 276 } }, "compatible_strategies": [ "single_node_tp", "single_node_tep", "single_node_dep", "multi_node_tp", "multi_node_tp_pp", "multi_node_tep", "multi_node_dep" ], "features": [ "tool_calling", "reasoning" ], "opt_in_features": [], "url": "/MiniMaxAI/MiniMax-M2.7", "json": "/MiniMaxAI/MiniMax-M2.7.json" }, { "hf_id": "MiniMaxAI/MiniMax-M2", "hf_org": "MiniMaxAI", "hf_repo": "MiniMax-M2", "title": "MiniMax-M2", "provider": "MiniMax", "description": "MiniMax M2 MoE language model (230B total / 10B active) for coding, agent toolchains, and long-context reasoning — native FP8 checkpoint", "date_updated": "2026-04-17", "difficulty": "intermediate", "min_vllm_version": "0.11.0", "architecture": "moe", "parameter_count": "230B", "active_parameters": "10B", "context_length": 196608, "tasks": [ "text" ], "performance_headline": "Open-source MoE with strong SWE-Bench and Terminal-Bench performance, 196K context", "variants": { "default": { "precision": "fp8", "vram_minimum_gb": 276 } }, "compatible_strategies": [ "single_node_tp", "single_node_tep", "single_node_dep", "multi_node_tp", "multi_node_tp_pp", "multi_node_tep", "multi_node_dep" ], "features": [ "tool_calling", "reasoning" ], "opt_in_features": [], "url": "/MiniMaxAI/MiniMax-M2", "json": "/MiniMaxAI/MiniMax-M2.json" }, { "hf_id": "OpenGVLab/InternVL3_5-8B", "hf_org": "OpenGVLab", "hf_repo": "InternVL3_5-8B", "title": "InternVL3.5", "provider": "InternVL (OpenGVLab)", "description": "InternVL 3.5 vision-language models from Shanghai AI Lab with thinking-mode prompting", "date_updated": "2026-04-17", "difficulty": "beginner", "min_vllm_version": "0.10.0", "architecture": "dense", "parameter_count": "8B", "active_parameters": "8B", "context_length": 40960, "tasks": [ "multimodal" ], "variants": { "default": { "precision": "bf16", "vram_minimum_gb": 19 } }, "compatible_strategies": [ "single_node_tp", "multi_node_tp", "multi_node_tp_pp" ], "features": [], "opt_in_features": [], "url": "/OpenGVLab/InternVL3_5-8B", "json": "/OpenGVLab/InternVL3_5-8B.json" }, { "hf_id": "PaddlePaddle/PaddleOCR-VL", "hf_org": "PaddlePaddle", "hf_repo": "PaddleOCR-VL", "title": "PaddleOCR-VL", "provider": "PaddlePaddle", "description": "PaddleOCR-VL (0.9B) — compact vision-language model for document parsing, OCR, tables, formulas, charts", "date_updated": "2026-04-17", "difficulty": "beginner", "min_vllm_version": "0.11.1", "architecture": "dense", "parameter_count": "0.9B", "active_parameters": "0.9B", "context_length": 131072, "tasks": [ "multimodal" ], "variants": { "default": { "precision": "bf16", "vram_minimum_gb": 2 } }, "compatible_strategies": [ "single_node_tp", "multi_node_tp", "multi_node_tp_pp" ], "features": [], "opt_in_features": [], "url": "/PaddlePaddle/PaddleOCR-VL", "json": "/PaddlePaddle/PaddleOCR-VL.json" }, { "hf_id": "Qwen/Qwen-Image", "hf_org": "Qwen", "hf_repo": "Qwen-Image", "title": "Qwen-Image", "provider": "Qwen", "description": "Text-to-image diffusion model (20B parameters) from the Qwen-Image family, served via vLLM-Omni.", "date_updated": "2026-04-17", "difficulty": "intermediate", "min_vllm_version": "0.18.0", "architecture": "dense", "parameter_count": "20B", "active_parameters": "20B", "context_length": 0, "tasks": [ "omni" ], "performance_headline": "Shared DiT core across T2I, image editing, and layered-image variants; accelerated via Cache-DiT, TeaCache, and sequence parallelism", "variants": { "default": { "precision": "bf16", "vram_minimum_gb": 48 }, "fp8": { "precision": "fp8", "vram_minimum_gb": 24 }, "int8": { "precision": "int8", "vram_minimum_gb": 24 } }, "compatible_strategies": [ "single_node_tp" ], "features": [], "opt_in_features": [], "url": "/Qwen/Qwen-Image", "json": "/Qwen/Qwen-Image.json" }, { "hf_id": "Qwen/Qwen2.5-VL-72B-Instruct", "hf_org": "Qwen", "hf_repo": "Qwen2.5-VL-72B-Instruct", "title": "Qwen2.5-VL-72B-Instruct", "provider": "Qwen", "description": "Qwen2.5-VL dense vision-language model (72B) for high-quality image and video understanding.", "date_updated": "2026-04-17", "difficulty": "intermediate", "min_vllm_version": "0.7.0", "architecture": "dense", "parameter_count": "72B", "active_parameters": "72B", "context_length": 128000, "tasks": [ "multimodal", "text" ], "performance_headline": "Verified on 4x A100 and 4x MI300X/MI325X/MI355X with BF16", "variants": { "default": { "precision": "bf16", "vram_minimum_gb": 173 }, "awq": { "precision": "int4", "vram_minimum_gb": 43 } }, "compatible_strategies": [ "single_node_tp", "multi_node_tp", "multi_node_tp_pp" ], "features": [], "opt_in_features": [], "url": "/Qwen/Qwen2.5-VL-72B-Instruct", "json": "/Qwen/Qwen2.5-VL-72B-Instruct.json" }, { "hf_id": "Qwen/Qwen3-235B-A22B-Instruct-2507", "hf_org": "Qwen", "hf_repo": "Qwen3-235B-A22B-Instruct-2507", "title": "Qwen3-235B-A22B-Instruct", "provider": "Qwen", "description": "Flagship Qwen3 MoE instruct model with 235B total and 22B active parameters, tuned for high-quality text generation.", "date_updated": "2026-04-17", "difficulty": "intermediate", "min_vllm_version": "0.10.0", "architecture": "moe", "parameter_count": "235B", "active_parameters": "22B", "context_length": 262144, "tasks": [ "text" ], "performance_headline": "Verified on 4x/8x H200, MI300X/MI325X/MI355X nodes (BF16 and FP8)", "variants": { "default": { "precision": "bf16", "vram_minimum_gb": 564 }, "fp8": { "precision": "fp8", "vram_minimum_gb": 240 }, "nvfp4": { "precision": "nvfp4", "vram_minimum_gb": 141 } }, "compatible_strategies": [ "single_node_tp", "single_node_tep", "single_node_dep", "multi_node_tp", "multi_node_tp_pp", "multi_node_dep", "multi_node_tep", "pd_cluster" ], "features": [ "tool_calling" ], "opt_in_features": [], "url": "/Qwen/Qwen3-235B-A22B-Instruct-2507", "json": "/Qwen/Qwen3-235B-A22B-Instruct-2507.json" }, { "hf_id": "Qwen/Qwen3-ASR-1.7B", "hf_org": "Qwen", "hf_repo": "Qwen3-ASR-1.7B", "title": "Qwen3-ASR-1.7B", "provider": "Qwen", "description": "Speech-to-text model supporting 11 languages, multiple accents, and singing voice with customizable text-context prompting.", "date_updated": "2026-04-17", "difficulty": "beginner", "min_vllm_version": "0.12.0", "architecture": "dense", "parameter_count": "2.3B", "active_parameters": "2.3B", "context_length": 65536, "tasks": [ "multimodal" ], "performance_headline": "Accurate multilingual ASR, including singing voice; single-GPU serving", "variants": { "default": { "precision": "bf16", "vram_minimum_gb": 4 } }, "compatible_strategies": [ "single_node_tp", "multi_node_tp", "multi_node_tp_pp" ], "features": [], "opt_in_features": [], "url": "/Qwen/Qwen3-ASR-1.7B", "json": "/Qwen/Qwen3-ASR-1.7B.json" }, { "hf_id": "Qwen/Qwen3-Coder-480B-A35B-Instruct", "hf_org": "Qwen", "hf_repo": "Qwen3-Coder-480B-A35B-Instruct", "title": "Qwen3-Coder-480B-A35B-Instruct", "provider": "Qwen", "description": "Large coder MoE with 480B total / 35B active parameters, strong tool-use and code generation capabilities.", "date_updated": "2026-04-17", "difficulty": "advanced", "min_vllm_version": "0.10.0", "architecture": "moe", "parameter_count": "480B", "active_parameters": "35B", "context_length": 262144, "tasks": [ "text" ], "performance_headline": "HumanEval 0.939, MBPP 0.918 (FP8). Recommended FP8 on 8x H200/H20 via DP=8", "variants": { "default": { "precision": "bf16", "vram_minimum_gb": 1152 }, "fp8": { "precision": "fp8", "vram_minimum_gb": 576 }, "nvfp4": { "precision": "nvfp4", "vram_minimum_gb": 288 } }, "compatible_strategies": [ "single_node_tp", "single_node_tep", "single_node_dep", "multi_node_tp", "multi_node_tp_pp", "multi_node_dep", "multi_node_tep", "pd_cluster" ], "features": [ "tool_calling" ], "opt_in_features": [], "url": "/Qwen/Qwen3-Coder-480B-A35B-Instruct", "json": "/Qwen/Qwen3-Coder-480B-A35B-Instruct.json" }, { "hf_id": "Qwen/Qwen3-Next-80B-A3B-Instruct", "hf_org": "Qwen", "hf_repo": "Qwen3-Next-80B-A3B-Instruct", "title": "Qwen3-Next-80B-A3B-Instruct", "provider": "Qwen", "description": "Advanced Qwen3-Next MoE model (80B total / 3B active) with hybrid attention, highly sparse experts, and multi-token prediction.", "date_updated": "2026-04-17", "difficulty": "intermediate", "min_vllm_version": "0.10.0", "architecture": "moe", "parameter_count": "80B", "active_parameters": "3B", "context_length": 262144, "tasks": [ "text" ], "performance_headline": "Highly sparse MoE with MTP-accelerated decoding, runs on 4x H200/H20/A100/A800", "variants": { "default": { "precision": "bf16", "vram_minimum_gb": 192 }, "fp8": { "precision": "fp8", "vram_minimum_gb": 96 }, "nvfp4": { "precision": "nvfp4", "vram_minimum_gb": 48 } }, "compatible_strategies": [ "single_node_tp", "single_node_tep", "single_node_dep", "multi_node_tp", "multi_node_tp_pp", "multi_node_dep", "multi_node_tep" ], "features": [ "tool_calling", "spec_decoding" ], "opt_in_features": [ "spec_decoding" ], "url": "/Qwen/Qwen3-Next-80B-A3B-Instruct", "json": "/Qwen/Qwen3-Next-80B-A3B-Instruct.json" }, { "hf_id": "Qwen/Qwen3-VL-235B-A22B-Instruct", "hf_org": "Qwen", "hf_repo": "Qwen3-VL-235B-A22B-Instruct", "title": "Qwen3-VL-235B-A22B-Instruct", "provider": "Qwen", "description": "Qwen3-VL flagship MoE vision-language model with 235B total / 22B active parameters, supporting images, video, and long context.", "date_updated": "2026-04-17", "difficulty": "intermediate", "min_vllm_version": "0.11.0", "architecture": "moe", "parameter_count": "235B", "active_parameters": "22B", "context_length": 262144, "tasks": [ "multimodal", "text" ], "performance_headline": "Strong on images, video, and text — #1 open model on text on lmarena.ai at release", "variants": { "default": { "precision": "bf16", "vram_minimum_gb": 564 }, "fp8": { "precision": "fp8", "vram_minimum_gb": 282 }, "nvfp4": { "precision": "nvfp4", "vram_minimum_gb": 141 } }, "compatible_strategies": [ "single_node_tp", "single_node_tep", "single_node_dep", "multi_node_tp", "multi_node_tp_pp", "multi_node_dep", "multi_node_tep", "pd_cluster" ], "features": [], "opt_in_features": [], "url": "/Qwen/Qwen3-VL-235B-A22B-Instruct", "json": "/Qwen/Qwen3-VL-235B-A22B-Instruct.json" }, { "hf_id": "Qwen/Qwen3.5-397B-A17B", "hf_org": "Qwen", "hf_repo": "Qwen3.5-397B-A17B", "title": "Qwen3.5-397B", "provider": "Qwen", "description": "Multimodal MoE model with gated delta networks architecture, 397B total / 17B active parameters, up to 262K context", "date_updated": "2026-04-16", "difficulty": "intermediate", "min_vllm_version": "0.17.0", "architecture": "moe", "parameter_count": "397B", "active_parameters": "17B", "context_length": 262144, "tasks": [ "text" ], "performance_headline": "Verified on 8x H200, 8x MI300X/MI355X, and GB200 nodes", "variants": { "default": { "precision": "bf16", "vram_minimum_gb": 953 }, "nvfp4": { "precision": "nvfp4", "vram_minimum_gb": 238 } }, "compatible_strategies": [ "single_node_tp", "single_node_tep", "single_node_dep", "multi_node_tp", "multi_node_tp_pp", "multi_node_dep", "multi_node_tep", "pd_cluster" ], "features": [ "tool_calling", "reasoning", "spec_decoding" ], "opt_in_features": [ "spec_decoding" ], "url": "/Qwen/Qwen3.5-397B-A17B", "json": "/Qwen/Qwen3.5-397B-A17B.json" }, { "hf_id": "Qwen/Qwen3.6-35B-A3B", "hf_org": "Qwen", "hf_repo": "Qwen3.6-35B-A3B", "title": "Qwen3.6-35B-A3B", "provider": "Qwen", "description": "Smaller Qwen3.6 multimodal MoE model (35B total / 3B active) with 256 experts (8 routed + 1 shared), gated delta networks architecture, and 262K context", "date_updated": "2026-04-18", "difficulty": "beginner", "min_vllm_version": "0.17.0", "architecture": "moe", "parameter_count": "35B", "active_parameters": "3B", "context_length": 262144, "tasks": [ "text" ], "performance_headline": "Compact Qwen3.6 MoE with 3B active parameters — single-GPU FP8 or 2-4 GPU BF16 serving", "variants": { "default": { "precision": "bf16", "vram_minimum_gb": 84 }, "fp8": { "precision": "fp8", "vram_minimum_gb": 42 } }, "compatible_strategies": [ "single_node_tp", "single_node_tep", "single_node_dep", "multi_node_tp", "multi_node_tp_pp", "multi_node_dep", "multi_node_tep" ], "features": [ "tool_calling", "reasoning", "spec_decoding" ], "opt_in_features": [ "spec_decoding" ], "url": "/Qwen/Qwen3.6-35B-A3B", "json": "/Qwen/Qwen3.6-35B-A3B.json" }, { "hf_id": "Qwen/Qwen3Guard-Gen-8B", "hf_org": "Qwen", "hf_repo": "Qwen3Guard-Gen-8B", "title": "Qwen3Guard-Gen-8B", "provider": "Qwen", "description": "Lightweight text-only guardrail/safety classifier model in the Qwen3Guard family.", "date_updated": "2026-04-17", "difficulty": "beginner", "min_vllm_version": "0.10.0", "architecture": "dense", "parameter_count": "8B", "active_parameters": "8B", "context_length": 32768, "tasks": [ "text" ], "performance_headline": "Runs on a single GPU; serves safety classifications over OpenAI-compatible API", "variants": { "default": { "precision": "bf16", "vram_minimum_gb": 19 }, "small_4b": { "precision": "bf16", "vram_minimum_gb": 10 }, "tiny_0_6b": { "precision": "bf16", "vram_minimum_gb": 4 } }, "compatible_strategies": [ "single_node_tp", "multi_node_tp", "multi_node_tp_pp" ], "features": [], "opt_in_features": [], "url": "/Qwen/Qwen3Guard-Gen-8B", "json": "/Qwen/Qwen3Guard-Gen-8B.json" }, { "hf_id": "Wan-AI/Wan2.2-T2V-A14B-Diffusers", "hf_org": "Wan-AI", "hf_repo": "Wan2.2-T2V-A14B-Diffusers", "title": "Wan2.2", "provider": "Wan AI", "description": "Wan2.2 video generation models — T2V/I2V MoE (14B active) and unified TI2V (5B dense), served via vLLM-Omni", "date_updated": "2026-04-17", "difficulty": "intermediate", "min_vllm_version": "0.12.0", "architecture": "moe", "parameter_count": "28B", "active_parameters": "14B", "context_length": 0, "tasks": [ "omni" ], "variants": { "default": { "precision": "bf16", "vram_minimum_gb": 152 }, "i2v": { "precision": "bf16", "vram_minimum_gb": 40 }, "ti2v_5b": { "precision": "bf16", "vram_minimum_gb": 20 } }, "compatible_strategies": [], "features": [], "opt_in_features": [], "url": "/Wan-AI/Wan2.2-T2V-A14B-Diffusers", "json": "/Wan-AI/Wan2.2-T2V-A14B-Diffusers.json" }, { "hf_id": "XiaomiMiMo/MiMo-V2-Flash", "hf_org": "XiaomiMiMo", "hf_repo": "MiMo-V2-Flash", "title": "MiMo-V2-Flash", "provider": "Xiaomi MiMo", "description": "Xiaomi's MoE reasoning model (309B total / 15B active) with hybrid attention and MTP for fast agentic workflows", "date_updated": "2026-04-17", "difficulty": "intermediate", "min_vllm_version": "0.11.0", "architecture": "moe", "parameter_count": "309B", "active_parameters": "15B", "context_length": 262144, "tasks": [ "text" ], "variants": { "default": { "precision": "fp8", "vram_minimum_gb": 371 } }, "compatible_strategies": [ "single_node_tp", "single_node_tep", "multi_node_tp", "multi_node_tp_pp", "multi_node_tep", "multi_node_dep" ], "features": [ "tool_calling", "reasoning" ], "opt_in_features": [], "url": "/XiaomiMiMo/MiMo-V2-Flash", "json": "/XiaomiMiMo/MiMo-V2-Flash.json" }, { "hf_id": "arcee-ai/Trinity-Large-Thinking", "hf_org": "arcee-ai", "hf_repo": "Trinity-Large-Thinking", "title": "Trinity-Large-Thinking", "provider": "Arcee AI", "description": "Arcee AI's reasoning-focused sparse MoE (AfmoeForCausalLM) with structured traces and agentic tool use", "date_updated": "2026-04-17", "difficulty": "intermediate", "min_vllm_version": "0.11.1", "architecture": "moe", "parameter_count": "398B", "active_parameters": "13B", "context_length": 262144, "tasks": [ "text" ], "variants": { "default": { "precision": "bf16", "vram_minimum_gb": 955 }, "nvfp4": { "precision": "nvfp4", "vram_minimum_gb": 239 } }, "compatible_strategies": [ "single_node_tp", "multi_node_tp", "multi_node_tp_pp", "multi_node_tep", "multi_node_dep" ], "features": [ "tool_calling", "reasoning" ], "opt_in_features": [], "url": "/arcee-ai/Trinity-Large-Thinking", "json": "/arcee-ai/Trinity-Large-Thinking.json" }, { "hf_id": "baidu/ERNIE-4.5-21B-A3B-PT", "hf_org": "baidu", "hf_repo": "ERNIE-4.5-21B-A3B-PT", "title": "ERNIE-4.5", "provider": "Ernie (Baidu)", "description": "Baidu ERNIE 4.5 MoE text models (21B-A3B, 300B-A47B) with BF16 and FP8 support plus ERNIE-MTP speculative decoding", "date_updated": "2026-04-17", "difficulty": "intermediate", "min_vllm_version": "0.10.1", "architecture": "moe", "parameter_count": "21B", "active_parameters": "3B", "context_length": 131072, "tasks": [ "text" ], "variants": { "default": { "precision": "bf16", "vram_minimum_gb": 106 }, "300b": { "precision": "bf16", "vram_minimum_gb": 640 } }, "compatible_strategies": [ "single_node_tp", "multi_node_tp", "multi_node_tp_pp", "multi_node_tep", "multi_node_dep" ], "features": [ "spec_decoding" ], "opt_in_features": [ "spec_decoding" ], "url": "/baidu/ERNIE-4.5-21B-A3B-PT", "json": "/baidu/ERNIE-4.5-21B-A3B-PT.json" }, { "hf_id": "baidu/ERNIE-4.5-VL-28B-A3B-PT", "hf_org": "baidu", "hf_repo": "ERNIE-4.5-VL-28B-A3B-PT", "title": "ERNIE-4.5-VL", "provider": "Ernie (Baidu)", "description": "Baidu ERNIE 4.5 VL MoE vision-language models (28B-A3B, 424B-A47B) with heterogeneous text/vision experts", "date_updated": "2026-04-17", "difficulty": "intermediate", "min_vllm_version": "0.11.0", "architecture": "moe", "parameter_count": "28B", "active_parameters": "3B", "context_length": 131072, "tasks": [ "multimodal" ], "variants": { "default": { "precision": "bf16", "vram_minimum_gb": 67 }, "424b": { "precision": "bf16", "vram_minimum_gb": 1120 }, "424b_fp8": { "precision": "fp8", "vram_minimum_gb": 640 } }, "compatible_strategies": [ "single_node_tp", "multi_node_tp", "multi_node_tp_pp", "multi_node_tep", "multi_node_dep" ], "features": [], "opt_in_features": [], "url": "/baidu/ERNIE-4.5-VL-28B-A3B-PT", "json": "/baidu/ERNIE-4.5-VL-28B-A3B-PT.json" }, { "hf_id": "deepseek-ai/DeepSeek-OCR-2", "hf_org": "deepseek-ai", "hf_repo": "DeepSeek-OCR-2", "title": "DeepSeek-OCR-2", "provider": "DeepSeek", "description": "Next-generation DeepSeek OCR model with improved document-to-markdown grounding and optical context compression.", "date_updated": "2026-04-17", "difficulty": "intermediate", "min_vllm_version": "0.12.0", "architecture": "moe", "parameter_count": "3B", "active_parameters": "3B", "context_length": 8192, "tasks": [ "multimodal" ], "performance_headline": "Improved grounding and markdown conversion over DeepSeek-OCR", "variants": { "default": { "precision": "bf16", "vram_minimum_gb": 7 } }, "compatible_strategies": [ "single_node_tp", "multi_node_tp", "multi_node_tp_pp", "multi_node_tep", "multi_node_dep" ], "features": [], "opt_in_features": [], "url": "/deepseek-ai/DeepSeek-OCR-2", "json": "/deepseek-ai/DeepSeek-OCR-2.json" }, { "hf_id": "deepseek-ai/DeepSeek-OCR", "hf_org": "deepseek-ai", "hf_repo": "DeepSeek-OCR", "title": "DeepSeek-OCR", "provider": "DeepSeek", "description": "Frontier OCR model exploring optical context compression for LLMs, optimized for document parsing and markdown generation.", "date_updated": "2026-04-17", "difficulty": "intermediate", "min_vllm_version": "0.12.0", "architecture": "moe", "parameter_count": "3B", "active_parameters": "3B", "context_length": 8192, "tasks": [ "multimodal" ], "performance_headline": "Optical context compression for efficient OCR and document understanding", "variants": { "default": { "precision": "bf16", "vram_minimum_gb": 7 } }, "compatible_strategies": [ "single_node_tp", "multi_node_tp", "multi_node_tp_pp", "multi_node_tep", "multi_node_dep" ], "features": [], "opt_in_features": [], "url": "/deepseek-ai/DeepSeek-OCR", "json": "/deepseek-ai/DeepSeek-OCR.json" }, { "hf_id": "deepseek-ai/DeepSeek-R1", "hf_org": "deepseek-ai", "hf_repo": "DeepSeek-R1", "title": "DeepSeek-R1", "provider": "DeepSeek", "description": "DeepSeek-R1 is a 671B-parameter MoE reasoning model built on the DeepSeek-V3 architecture, trained with large-scale reinforcement learning for strong chain-of-thought capabilities.", "date_updated": "2026-04-17", "difficulty": "intermediate", "min_vllm_version": "0.12.0", "architecture": "moe", "parameter_count": "671B", "active_parameters": "37B", "context_length": 163840, "tasks": [ "text" ], "performance_headline": "Open-weights RL-trained reasoning model with native FP8 / FP4 variants", "variants": { "default": { "precision": "fp8", "vram_minimum_gb": 805 }, "r1_0528": { "precision": "fp8", "vram_minimum_gb": 805 }, "nvfp4": { "precision": "nvfp4", "vram_minimum_gb": 403 } }, "compatible_strategies": [ "single_node_tp", "single_node_tep", "single_node_dep", "multi_node_tp", "multi_node_tp_pp", "multi_node_dep", "multi_node_tep", "pd_cluster" ], "features": [], "opt_in_features": [], "url": "/deepseek-ai/DeepSeek-R1", "json": "/deepseek-ai/DeepSeek-R1.json" }, { "hf_id": "deepseek-ai/DeepSeek-V3.1", "hf_org": "deepseek-ai", "hf_repo": "DeepSeek-V3.1", "title": "DeepSeek-V3.1", "provider": "DeepSeek", "description": "DeepSeek-V3.1 is a hybrid MoE model that supports dynamic switching between thinking and non-thinking modes, with tool calling and function execution.", "date_updated": "2026-04-17", "difficulty": "intermediate", "min_vllm_version": "0.12.0", "architecture": "moe", "parameter_count": "671B", "active_parameters": "37B", "context_length": 163840, "tasks": [ "text" ], "performance_headline": "Hybrid thinking / non-thinking MoE with native FP8 and tool calling", "variants": { "default": { "precision": "fp8", "vram_minimum_gb": 805 }, "nvfp4": { "precision": "nvfp4", "vram_minimum_gb": 403 } }, "compatible_strategies": [ "single_node_tp", "single_node_tep", "single_node_dep", "multi_node_tp", "multi_node_tp_pp", "multi_node_dep", "multi_node_tep", "pd_cluster" ], "features": [ "tool_calling", "reasoning" ], "opt_in_features": [], "url": "/deepseek-ai/DeepSeek-V3.1", "json": "/deepseek-ai/DeepSeek-V3.1.json" }, { "hf_id": "deepseek-ai/DeepSeek-V3.2-Exp", "hf_org": "deepseek-ai", "hf_repo": "DeepSeek-V3.2-Exp", "title": "DeepSeek-V3.2-Exp", "provider": "DeepSeek", "description": "Experimental DeepSeek-V3.2 preview with sparse attention (MQA-like logits) and FP8 KV cache; architecture matches DeepSeek-V3.1 except for the sparse attention mechanism.", "date_updated": "2026-04-17", "difficulty": "advanced", "min_vllm_version": "0.12.0", "architecture": "moe", "parameter_count": "671B", "active_parameters": "37B", "context_length": 163840, "tasks": [ "text" ], "performance_headline": "Sparse attention MoE with FP8 KV cache and strong GSM8K score (~0.96)", "variants": { "default": { "precision": "fp8", "vram_minimum_gb": 805 } }, "compatible_strategies": [ "single_node_tp", "single_node_tep", "single_node_dep", "multi_node_tp", "multi_node_tp_pp", "multi_node_dep", "multi_node_tep" ], "features": [ "reasoning" ], "opt_in_features": [], "url": "/deepseek-ai/DeepSeek-V3.2-Exp", "json": "/deepseek-ai/DeepSeek-V3.2-Exp.json" }, { "hf_id": "deepseek-ai/DeepSeek-V3.2", "hf_org": "deepseek-ai", "hf_repo": "DeepSeek-V3.2", "title": "DeepSeek-V3.2", "provider": "DeepSeek", "description": "DeepSeek V3.2 MoE model with MLA attention, sparse attention, and scalable RL for strong reasoning and agent capabilities.", "date_updated": "2026-04-01", "difficulty": "intermediate", "min_vllm_version": "0.18.0", "architecture": "moe", "parameter_count": "671B", "active_parameters": "37B", "context_length": 163840, "tasks": [ "text" ], "performance_headline": "GPT-5-level reasoning with efficient MoE inference", "variants": { "default": { "precision": "fp8", "vram_minimum_gb": 805 }, "nvfp4": { "precision": "nvfp4", "vram_minimum_gb": 403 } }, "compatible_strategies": [ "single_node_tp", "single_node_tep", "single_node_dep", "multi_node_tp", "multi_node_tp_pp", "multi_node_dep", "multi_node_tep", "pd_cluster" ], "features": [ "tool_calling", "reasoning", "spec_decoding" ], "opt_in_features": [ "spec_decoding" ], "url": "/deepseek-ai/DeepSeek-V3.2", "json": "/deepseek-ai/DeepSeek-V3.2.json" }, { "hf_id": "deepseek-ai/DeepSeek-V3", "hf_org": "deepseek-ai", "hf_repo": "DeepSeek-V3", "title": "DeepSeek-V3", "provider": "DeepSeek", "description": "DeepSeek-V3 is a 671B-parameter Mixture-of-Experts model with native FP8 weights and strong reasoning, coding, and math capabilities.", "date_updated": "2026-04-17", "difficulty": "intermediate", "min_vllm_version": "0.12.0", "architecture": "moe", "parameter_count": "671B", "active_parameters": "37B", "context_length": 163840, "tasks": [ "text" ], "performance_headline": "Frontier open-weights MoE with native FP8 and FP4 variants", "variants": { "default": { "precision": "fp8", "vram_minimum_gb": 805 }, "fp4": { "precision": "fp4", "vram_minimum_gb": 403 } }, "compatible_strategies": [ "single_node_tp", "single_node_tep", "single_node_dep", "multi_node_tp", "multi_node_tp_pp", "multi_node_dep", "multi_node_tep", "pd_cluster" ], "features": [], "opt_in_features": [], "url": "/deepseek-ai/DeepSeek-V3", "json": "/deepseek-ai/DeepSeek-V3.json" }, { "hf_id": "inclusionAI/Ring-1T-FP8", "hf_org": "inclusionAI", "hf_repo": "Ring-1T-FP8", "title": "Ring-1T-FP8", "provider": "inclusionAI", "description": "Ring-1T (BailingMoeV2) FP8 model (~1T total params) for 8xH200 or 8xMI300X deployment", "date_updated": "2026-04-17", "difficulty": "advanced", "min_vllm_version": "0.11.0", "architecture": "moe", "parameter_count": "1T", "active_parameters": "50B", "context_length": 65536, "tasks": [ "text" ], "variants": { "default": { "precision": "fp8", "vram_minimum_gb": 1200 } }, "compatible_strategies": [ "single_node_tp", "multi_node_tp", "multi_node_tp_pp", "multi_node_tep", "multi_node_dep" ], "features": [], "opt_in_features": [], "url": "/inclusionAI/Ring-1T-FP8", "json": "/inclusionAI/Ring-1T-FP8.json" }, { "hf_id": "internlm/Intern-S1", "hf_org": "internlm", "hf_repo": "Intern-S1", "title": "Intern-S1", "provider": "InternLM", "description": "Intern-S1 vision-language model from Shanghai AI Lab with BF16/FP8 variants and thinking/non-thinking modes", "date_updated": "2026-04-17", "difficulty": "intermediate", "min_vllm_version": "0.10.0", "architecture": "moe", "parameter_count": "241B", "active_parameters": "28B", "context_length": 65536, "tasks": [ "multimodal" ], "variants": { "default": { "precision": "bf16", "vram_minimum_gb": 578 }, "fp8": { "precision": "fp8", "vram_minimum_gb": 289 } }, "compatible_strategies": [ "single_node_tp", "multi_node_tp", "multi_node_tp_pp", "multi_node_tep", "multi_node_dep" ], "features": [ "tool_calling", "reasoning" ], "opt_in_features": [], "url": "/internlm/Intern-S1", "json": "/internlm/Intern-S1.json" }, { "hf_id": "jinaai/jina-reranker-m0", "hf_org": "jinaai", "hf_repo": "jina-reranker-m0", "title": "Jina Reranker m0", "provider": "Jina AI", "description": "Multilingual, multimodal reranker for text and visual documents across 29+ languages via Qwen2-VL backbone", "date_updated": "2026-04-17", "difficulty": "beginner", "min_vllm_version": "0.8.0", "architecture": "dense", "parameter_count": "2.4B", "active_parameters": "2.4B", "context_length": 32768, "tasks": [ "embedding" ], "variants": { "default": { "precision": "bf16", "vram_minimum_gb": 6 } }, "compatible_strategies": [ "single_node_tp", "multi_node_tp", "multi_node_tp_pp" ], "features": [], "opt_in_features": [], "url": "/jinaai/jina-reranker-m0", "json": "/jinaai/jina-reranker-m0.json" }, { "hf_id": "meituan-longcat/LongCat-Image-Edit", "hf_org": "meituan-longcat", "hf_repo": "LongCat-Image-Edit", "title": "LongCat-Image-Edit", "provider": "Meituan", "description": "Bilingual (Chinese-English) image editing model from Meituan LongCat, served via vLLM-Omni", "date_updated": "2026-04-17", "difficulty": "intermediate", "min_vllm_version": "0.12.0", "architecture": "dense", "parameter_count": "6B", "active_parameters": "6B", "context_length": 0, "tasks": [ "omni" ], "variants": { "default": { "precision": "bf16", "vram_minimum_gb": 36 } }, "compatible_strategies": [], "features": [], "opt_in_features": [], "url": "/meituan-longcat/LongCat-Image-Edit", "json": "/meituan-longcat/LongCat-Image-Edit.json" }, { "hf_id": "meta-llama/Llama-3.1-8B-Instruct", "hf_org": "meta-llama", "hf_repo": "Llama-3.1-8B-Instruct", "title": "Llama-3.1-8B-Instruct", "provider": "Meta", "description": "Meta's Llama 3.1 8B dense instruction-tuned language model with 128K context", "date_updated": "2026-04-17", "difficulty": "beginner", "min_vllm_version": "0.6.0", "architecture": "dense", "parameter_count": "8B", "active_parameters": "8B", "context_length": 131072, "tasks": [ "text" ], "variants": { "default": { "precision": "bf16", "vram_minimum_gb": 20 }, "nvfp4": { "precision": "nvfp4", "vram_minimum_gb": 5 }, "nvidia_fp8": { "precision": "fp8", "vram_minimum_gb": 10 } }, "compatible_strategies": [ "single_node_tp", "multi_node_tp", "multi_node_tp_pp" ], "features": [], "opt_in_features": [], "url": "/meta-llama/Llama-3.1-8B-Instruct", "json": "/meta-llama/Llama-3.1-8B-Instruct.json" }, { "hf_id": "meta-llama/Llama-3.3-70B-Instruct", "hf_org": "meta-llama", "hf_repo": "Llama-3.3-70B-Instruct", "title": "Llama-3.3-70B", "provider": "Meta", "description": "Llama 3.3 70B dense model with NVIDIA FP8/FP4 quantized variants for Hopper and Blackwell GPUs", "date_updated": "2026-04-16", "difficulty": "beginner", "min_vllm_version": "0.12.0", "architecture": "dense", "parameter_count": "70B", "active_parameters": "70B", "context_length": 131072, "tasks": [ "text" ], "performance_headline": "", "variants": { "default": { "precision": "bf16", "vram_minimum_gb": 170 }, "fp8": { "precision": "fp8", "vram_minimum_gb": 84 }, "nvfp4": { "precision": "nvfp4", "vram_minimum_gb": 42 } }, "compatible_strategies": [ "single_node_tp", "multi_node_tp", "multi_node_tp_pp" ], "features": [], "opt_in_features": [], "url": "/meta-llama/Llama-3.3-70B-Instruct", "json": "/meta-llama/Llama-3.3-70B-Instruct.json" }, { "hf_id": "meta-llama/Llama-4-Scout-17B-16E-Instruct", "hf_org": "meta-llama", "hf_repo": "Llama-4-Scout-17B-16E-Instruct", "title": "Llama-4-Scout", "provider": "Meta", "description": "Llama 4 Scout 17B-16E MoE model with NVIDIA FP8/FP4 variants, fits on a single GPU with quantization", "date_updated": "2026-04-16", "difficulty": "beginner", "min_vllm_version": "0.12.0", "architecture": "moe", "parameter_count": "109B", "active_parameters": "17B", "context_length": 10485760, "tasks": [ "text" ], "performance_headline": "", "variants": { "default": { "precision": "bf16", "vram_minimum_gb": 262 }, "fp8": { "precision": "fp8", "vram_minimum_gb": 131 }, "nvfp4": { "precision": "nvfp4", "vram_minimum_gb": 65 } }, "compatible_strategies": [ "single_node_tp", "single_node_tep", "single_node_dep", "multi_node_tp", "multi_node_tp_pp", "multi_node_tep", "multi_node_dep" ], "features": [], "opt_in_features": [], "url": "/meta-llama/Llama-4-Scout-17B-16E-Instruct", "json": "/meta-llama/Llama-4-Scout-17B-16E-Instruct.json" }, { "hf_id": "microsoft/Phi-4-mini-instruct", "hf_org": "microsoft", "hf_repo": "Phi-4-mini-instruct", "title": "Phi-4", "provider": "Microsoft", "description": "Microsoft's Phi-4 family of lightweight dense models (mini-instruct, reasoning, multimodal) with 128K context", "date_updated": "2026-04-17", "difficulty": "beginner", "min_vllm_version": "0.7.0", "architecture": "dense", "parameter_count": "4B", "active_parameters": "4B", "context_length": 131072, "tasks": [ "text", "multimodal" ], "variants": { "default": { "precision": "bf16", "vram_minimum_gb": 10 }, "mini_reasoning": { "precision": "bf16", "vram_minimum_gb": 10 }, "reasoning": { "precision": "bf16", "vram_minimum_gb": 30 }, "multimodal": { "precision": "bf16", "vram_minimum_gb": 16 } }, "compatible_strategies": [ "single_node_tp", "multi_node_tp", "multi_node_tp_pp" ], "features": [], "opt_in_features": [], "url": "/microsoft/Phi-4-mini-instruct", "json": "/microsoft/Phi-4-mini-instruct.json" }, { "hf_id": "mistralai/Ministral-3-14B-Instruct-2512", "hf_org": "mistralai", "hf_repo": "Ministral-3-14B-Instruct-2512", "title": "Ministral-3-Instruct", "provider": "Mistral AI", "description": "Ministral 3 Instruct family (3B/8B/14B) with FP8 weights, vision support, and 256K context", "date_updated": "2026-04-17", "difficulty": "beginner", "min_vllm_version": "0.11.0", "architecture": "dense", "parameter_count": "14B", "active_parameters": "14B", "context_length": 262144, "tasks": [ "multimodal" ], "variants": { "default": { "precision": "fp8", "vram_minimum_gb": 17 }, "8b": { "precision": "fp8", "vram_minimum_gb": 12 }, "3b": { "precision": "fp8", "vram_minimum_gb": 6 }, "fp8": { "precision": "fp8", "vram_minimum_gb": 17 } }, "compatible_strategies": [ "single_node_tp", "multi_node_tp", "multi_node_tp_pp" ], "features": [ "tool_calling" ], "opt_in_features": [], "url": "/mistralai/Ministral-3-14B-Instruct-2512", "json": "/mistralai/Ministral-3-14B-Instruct-2512.json" }, { "hf_id": "mistralai/Ministral-3-8B-Reasoning-2512", "hf_org": "mistralai", "hf_repo": "Ministral-3-8B-Reasoning-2512", "title": "Ministral-3-Reasoning", "provider": "Mistral AI", "description": "Ministral 3 Reasoning family (3B/8B/14B) with BF16 weights, vision support, and 256K context", "date_updated": "2026-04-17", "difficulty": "intermediate", "min_vllm_version": "0.11.0", "architecture": "dense", "parameter_count": "8B", "active_parameters": "8B", "context_length": 262144, "tasks": [ "multimodal" ], "variants": { "default": { "precision": "bf16", "vram_minimum_gb": 22 }, "3b": { "precision": "bf16", "vram_minimum_gb": 8 }, "14b": { "precision": "bf16", "vram_minimum_gb": 32 } }, "compatible_strategies": [ "single_node_tp", "multi_node_tp", "multi_node_tp_pp" ], "features": [ "tool_calling", "reasoning" ], "opt_in_features": [], "url": "/mistralai/Ministral-3-8B-Reasoning-2512", "json": "/mistralai/Ministral-3-8B-Reasoning-2512.json" }, { "hf_id": "mistralai/Mistral-Large-3-675B-Instruct-2512", "hf_org": "mistralai", "hf_repo": "Mistral-Large-3-675B-Instruct-2512", "title": "Mistral-Large-3-675B-Instruct", "provider": "Mistral AI", "description": "Mistral Large 3 (675B) with FP8 and NVFP4 weights for 8xH200 / 4xB200 deployments", "date_updated": "2026-04-17", "difficulty": "advanced", "min_vllm_version": "0.11.0", "architecture": "moe", "parameter_count": "675B", "active_parameters": "22B", "context_length": 294912, "tasks": [ "multimodal" ], "variants": { "default": { "precision": "fp8", "vram_minimum_gb": 810 }, "nvfp4": { "precision": "nvfp4", "vram_minimum_gb": 405 }, "fp8": { "precision": "fp8", "vram_minimum_gb": 810 } }, "compatible_strategies": [ "single_node_tp", "multi_node_tp", "multi_node_tp_pp", "multi_node_tep", "multi_node_dep" ], "features": [ "tool_calling" ], "opt_in_features": [], "url": "/mistralai/Mistral-Large-3-675B-Instruct-2512", "json": "/mistralai/Mistral-Large-3-675B-Instruct-2512.json" }, { "hf_id": "moonshotai/Kimi-K2-Instruct", "hf_org": "moonshotai", "hf_repo": "Kimi-K2-Instruct", "title": "Kimi-K2-Instruct", "provider": "Moonshot AI", "description": "Moonshot AI's Kimi-K2 is a trillion-parameter MoE instruction model (~32B active) with native FP8 weights and strong tool-calling capabilities.", "date_updated": "2026-04-17", "difficulty": "advanced", "min_vllm_version": "0.12.0", "architecture": "moe", "parameter_count": "1T", "active_parameters": "32B", "context_length": 131072, "tasks": [ "text" ], "performance_headline": "Open-weights 1T-parameter MoE with native FP8 and Kimi K2 tool calling", "variants": { "default": { "precision": "fp8", "vram_minimum_gb": 1200 } }, "compatible_strategies": [ "single_node_tp", "single_node_tep", "single_node_dep", "multi_node_tp", "multi_node_tp_pp", "multi_node_dep", "multi_node_tep", "pd_cluster" ], "features": [ "tool_calling" ], "opt_in_features": [], "url": "/moonshotai/Kimi-K2-Instruct", "json": "/moonshotai/Kimi-K2-Instruct.json" }, { "hf_id": "moonshotai/Kimi-K2-Thinking", "hf_org": "moonshotai", "hf_repo": "Kimi-K2-Thinking", "title": "Kimi-K2-Thinking", "provider": "Moonshot AI", "description": "Kimi-K2-Thinking is an advanced reasoning MoE model with native INT4 QAT weights, designed for long-horizon agent workflows interleaving chain-of-thought reasoning with tool calls.", "date_updated": "2026-04-17", "difficulty": "advanced", "min_vllm_version": "0.12.0", "architecture": "moe", "parameter_count": "1T", "active_parameters": "32B", "context_length": 262144, "tasks": [ "text" ], "performance_headline": "1T MoE thinking model with native INT4 QAT for 2x low-latency speed-up", "variants": { "default": { "precision": "int4", "vram_minimum_gb": 600 }, "nvfp4": { "precision": "nvfp4", "vram_minimum_gb": 600 } }, "compatible_strategies": [ "single_node_tp", "single_node_tep", "single_node_dep", "multi_node_tp", "multi_node_tp_pp", "multi_node_dep", "multi_node_tep", "pd_cluster" ], "features": [ "tool_calling", "reasoning" ], "opt_in_features": [], "url": "/moonshotai/Kimi-K2-Thinking", "json": "/moonshotai/Kimi-K2-Thinking.json" }, { "hf_id": "moonshotai/Kimi-K2.5", "hf_org": "moonshotai", "hf_repo": "Kimi-K2.5", "title": "Kimi-K2.5", "provider": "Moonshot AI", "description": "Open-source native multimodal agentic MoE model with vision-language understanding, tool calling, and thinking modes", "date_updated": "2026-04-16", "difficulty": "intermediate", "min_vllm_version": "0.15.0", "architecture": "moe", "parameter_count": "1T", "active_parameters": "32B", "context_length": 262144, "tasks": [ "multimodal" ], "performance_headline": "Multimodal agentic MoE model with DeepSeek-V3 backbone and MLA attention", "variants": { "default": { "precision": "int4", "vram_minimum_gb": 714 }, "nvfp4": { "precision": "nvfp4", "vram_minimum_gb": 600 } }, "compatible_strategies": [ "single_node_tp", "single_node_tep", "single_node_dep", "multi_node_tp", "multi_node_tp_pp", "multi_node_dep", "multi_node_tep", "pd_cluster" ], "features": [ "tool_calling", "reasoning", "spec_decoding" ], "opt_in_features": [ "spec_decoding" ], "url": "/moonshotai/Kimi-K2.5", "json": "/moonshotai/Kimi-K2.5.json" }, { "hf_id": "moonshotai/Kimi-Linear-48B-A3B-Instruct", "hf_org": "moonshotai", "hf_repo": "Kimi-Linear-48B-A3B-Instruct", "title": "Kimi-Linear-48B-A3B-Instruct", "provider": "Moonshot AI", "description": "Kimi-Linear is a 48B-parameter instruction-tuned MoE model (~3B activated) with a linear-attention variant supporting very long context (1M tokens).", "date_updated": "2026-04-17", "difficulty": "intermediate", "min_vllm_version": "0.11.2", "architecture": "moe", "parameter_count": "48B", "active_parameters": "3B", "context_length": 1048576, "tasks": [ "text" ], "performance_headline": "Linear-attention MoE with 1M-token context on a single node", "variants": { "default": { "precision": "bf16", "vram_minimum_gb": 115 } }, "compatible_strategies": [ "single_node_tp", "multi_node_tp", "multi_node_tp_pp", "multi_node_tep", "multi_node_dep" ], "features": [], "opt_in_features": [], "url": "/moonshotai/Kimi-Linear-48B-A3B-Instruct", "json": "/moonshotai/Kimi-Linear-48B-A3B-Instruct.json" }, { "hf_id": "nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-BF16", "hf_org": "nvidia", "hf_repo": "NVIDIA-Nemotron-3-Nano-30B-A3B-BF16", "title": "NVIDIA Nemotron-3-Nano-30B-A3B", "provider": "NVIDIA", "description": "NVIDIA Nemotron-3-Nano Mamba-hybrid MoE (30B total / ~3B active) with BF16 and FP8 variants", "date_updated": "2026-04-17", "difficulty": "intermediate", "min_vllm_version": "0.11.2", "architecture": "moe", "parameter_count": "30B", "active_parameters": "3B", "context_length": 262144, "tasks": [ "text" ], "variants": { "default": { "precision": "bf16", "vram_minimum_gb": 72 }, "fp8": { "precision": "fp8", "vram_minimum_gb": 35 } }, "compatible_strategies": [ "single_node_tp", "multi_node_tp", "multi_node_tp_pp", "multi_node_tep", "multi_node_dep" ], "features": [ "tool_calling", "reasoning" ], "opt_in_features": [], "url": "/nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-BF16", "json": "/nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-BF16.json" }, { "hf_id": "nvidia/NVIDIA-Nemotron-Nano-12B-v2-VL-BF16", "hf_org": "nvidia", "hf_repo": "NVIDIA-Nemotron-Nano-12B-v2-VL-BF16", "title": "NVIDIA Nemotron-Nano-12B-v2-VL", "provider": "NVIDIA", "description": "NVIDIA Nemotron-Nano 12B vision-language model with video support and Efficient Video Sampling (EVS)", "date_updated": "2026-04-17", "difficulty": "intermediate", "min_vllm_version": "0.11.1", "architecture": "dense", "parameter_count": "12B", "active_parameters": "12B", "context_length": 131072, "tasks": [ "multimodal" ], "variants": { "default": { "precision": "bf16", "vram_minimum_gb": 29 }, "fp8": { "precision": "fp8", "vram_minimum_gb": 14 }, "nvfp4": { "precision": "nvfp4", "vram_minimum_gb": 8 } }, "compatible_strategies": [ "single_node_tp", "multi_node_tp", "multi_node_tp_pp" ], "features": [ "video_compression" ], "opt_in_features": [], "url": "/nvidia/NVIDIA-Nemotron-Nano-12B-v2-VL-BF16", "json": "/nvidia/NVIDIA-Nemotron-Nano-12B-v2-VL-BF16.json" }, { "hf_id": "openai/gpt-oss-120b", "hf_org": "openai", "hf_repo": "gpt-oss-120b", "title": "GPT-OSS", "provider": "OpenAI", "description": "OpenAI's gpt-oss family (20B / 120B) with MXFP4 MoE, attention-sinks, built-in tools via Responses API", "date_updated": "2026-04-17", "difficulty": "intermediate", "min_vllm_version": "0.10.0", "architecture": "moe", "parameter_count": "120B", "active_parameters": "5.1B", "context_length": 131072, "tasks": [ "text" ], "variants": { "default": { "precision": "mxfp4", "vram_minimum_gb": 96 }, "20b": { "precision": "mxfp4", "vram_minimum_gb": 40 }, "amd_fp8": { "precision": "mxfp4", "vram_minimum_gb": 80 } }, "compatible_strategies": [ "single_node_tp", "multi_node_tp", "multi_node_tp_pp", "multi_node_tep", "multi_node_dep" ], "features": [ "tool_calling", "spec_decoding" ], "opt_in_features": [ "spec_decoding" ], "url": "/openai/gpt-oss-120b", "json": "/openai/gpt-oss-120b.json" }, { "hf_id": "stabilityai/stable-audio-open-1.0", "hf_org": "stabilityai", "hf_repo": "stable-audio-open-1.0", "title": "Stable Audio Open", "provider": "Stability AI", "description": "Text-to-audio generation model (1.2B params) producing up to ~47 s stereo audio at 44.1 kHz, served via vLLM-Omni", "date_updated": "2026-04-17", "difficulty": "beginner", "min_vllm_version": "0.14.1", "architecture": "dense", "parameter_count": "1.2B", "active_parameters": "1.2B", "context_length": 0, "tasks": [ "omni" ], "variants": { "default": { "precision": "bf16", "vram_minimum_gb": 19 } }, "compatible_strategies": [], "features": [], "opt_in_features": [], "url": "/stabilityai/stable-audio-open-1.0", "json": "/stabilityai/stable-audio-open-1.0.json" }, { "hf_id": "stabilityai/stable-diffusion-3.5-medium", "hf_org": "stabilityai", "hf_repo": "stable-diffusion-3.5-medium", "title": "Stable Diffusion 3.5", "provider": "Stability AI", "description": "Stability AI's Stable Diffusion 3.5 text-to-image family (medium 2.5B, large 8.1B, large-turbo) via vLLM-Omni with Cache-DiT acceleration", "date_updated": "2026-04-17", "difficulty": "beginner", "min_vllm_version": "0.12.0", "architecture": "dense", "parameter_count": "2.5B", "active_parameters": "2.5B", "context_length": 0, "tasks": [ "omni" ], "variants": { "default": { "precision": "bf16", "vram_minimum_gb": 44 }, "large": { "precision": "bf16", "vram_minimum_gb": 24 }, "large_turbo": { "precision": "bf16", "vram_minimum_gb": 24 } }, "compatible_strategies": [], "features": [], "opt_in_features": [], "url": "/stabilityai/stable-diffusion-3.5-medium", "json": "/stabilityai/stable-diffusion-3.5-medium.json" }, { "hf_id": "stepfun-ai/Step-3.5-Flash", "hf_org": "stepfun-ai", "hf_repo": "Step-3.5-Flash", "title": "Step-3.5-Flash", "provider": "StepFun", "description": "Production-grade reasoning MoE (~196B total / 11B active parameters) with hybrid attention schedules, SWA compensation, and multi-token prediction for low-latency long-context inference", "date_updated": "2026-04-17", "difficulty": "intermediate", "min_vllm_version": "0.11.0", "architecture": "moe", "parameter_count": "196B", "active_parameters": "11B", "context_length": 262144, "tasks": [ "text" ], "performance_headline": "Sparse MoE reasoning model with hybrid attention and step3p5 MTP speculative decoding", "variants": { "default": { "precision": "bf16", "vram_minimum_gb": 470 }, "fp8": { "precision": "fp8", "vram_minimum_gb": 235 }, "int4": { "precision": "int4", "vram_minimum_gb": 118 }, "int8": { "precision": "int8", "vram_minimum_gb": 235 } }, "compatible_strategies": [ "single_node_tp", "single_node_tep", "single_node_dep", "multi_node_tp", "multi_node_tp_pp", "multi_node_tep", "multi_node_dep" ], "features": [ "tool_calling", "reasoning", "spec_decoding" ], "opt_in_features": [ "spec_decoding" ], "url": "/stepfun-ai/Step-3.5-Flash", "json": "/stepfun-ai/Step-3.5-Flash.json" }, { "hf_id": "tencent/Hunyuan-A13B-Instruct", "hf_org": "tencent", "hf_repo": "Hunyuan-A13B-Instruct", "title": "Hunyuan-A13B-Instruct", "provider": "Tencent Hunyuan", "description": "Tencent Hunyuan A13B instruct-tuned MoE language model with AITER-accelerated AMD ROCm deployment", "date_updated": "2026-04-17", "difficulty": "beginner", "min_vllm_version": "0.11.0", "architecture": "moe", "parameter_count": "80B", "active_parameters": "13B", "context_length": 32768, "tasks": [ "text" ], "performance_headline": "Hunyuan-A13B MoE with AITER acceleration on AMD MI300X/MI325X/MI355X", "variants": { "default": { "precision": "bf16", "vram_minimum_gb": 580 }, "fp8": { "precision": "fp8", "vram_minimum_gb": 96 } }, "compatible_strategies": [ "single_node_tp", "single_node_tep", "multi_node_tp", "multi_node_tp_pp", "multi_node_tep", "multi_node_dep" ], "features": [], "opt_in_features": [], "url": "/tencent/Hunyuan-A13B-Instruct", "json": "/tencent/Hunyuan-A13B-Instruct.json" }, { "hf_id": "tencent/HunyuanOCR", "hf_org": "tencent", "hf_repo": "HunyuanOCR", "title": "HunyuanOCR", "provider": "Tencent Hunyuan", "description": "Tencent Hunyuan end-to-end OCR expert VLM (~1B) for online OCR serving with an OpenAI-compatible API", "date_updated": "2026-04-17", "difficulty": "beginner", "min_vllm_version": "0.11.0", "architecture": "dense", "parameter_count": "1B", "active_parameters": "1B", "context_length": 32768, "tasks": [ "multimodal" ], "performance_headline": "Compact 1B end-to-end OCR VLM from the Hunyuan native multimodal family", "variants": { "default": { "precision": "bf16", "vram_minimum_gb": 2 } }, "compatible_strategies": [ "single_node_tp", "multi_node_tp", "multi_node_tp_pp" ], "features": [], "opt_in_features": [], "url": "/tencent/HunyuanOCR", "json": "/tencent/HunyuanOCR.json" }, { "hf_id": "zai-org/GLM-4.5", "hf_org": "zai-org", "hf_repo": "GLM-4.5", "title": "GLM-4.5", "provider": "GLM (Z-AI)", "description": "GLM-4.5 MoE language model (~358B total parameters, BF16) with built-in MTP layers for speculative decoding and native tool calling", "date_updated": "2026-04-17", "difficulty": "intermediate", "min_vllm_version": "0.11.0", "architecture": "moe", "parameter_count": "358B", "active_parameters": "32B", "context_length": 131072, "tasks": [ "text" ], "performance_headline": "GLM-4.X series MoE model with native FP8 and BF16 support and MTP speculative decoding", "variants": { "default": { "precision": "bf16", "vram_minimum_gb": 859 }, "fp8": { "precision": "fp8", "vram_minimum_gb": 430 } }, "compatible_strategies": [ "single_node_tp", "single_node_tep", "single_node_dep", "multi_node_tp", "multi_node_tp_pp", "multi_node_dep", "multi_node_tep" ], "features": [ "tool_calling", "reasoning", "spec_decoding" ], "opt_in_features": [ "spec_decoding" ], "url": "/zai-org/GLM-4.5", "json": "/zai-org/GLM-4.5.json" }, { "hf_id": "zai-org/GLM-4.5V", "hf_org": "zai-org", "hf_repo": "GLM-4.5V", "title": "GLM-4.5V", "provider": "GLM (Z-AI)", "description": "GLM-4.5 vision-language MoE model (~107B parameters, BF16) with image-text-to-text capability, 64K context, expert parallelism, and native FP8", "date_updated": "2026-04-17", "difficulty": "intermediate", "min_vllm_version": "0.12.0", "architecture": "moe", "parameter_count": "107B", "active_parameters": "12B", "context_length": 65536, "tasks": [ "multimodal" ], "performance_headline": "Multimodal GLM-4.5V with native FP8 and expert parallelism, deploys on 4xH100", "variants": { "default": { "precision": "bf16", "vram_minimum_gb": 257 }, "fp8": { "precision": "fp8", "vram_minimum_gb": 128 } }, "compatible_strategies": [ "single_node_tp", "single_node_tep", "single_node_dep", "multi_node_tp", "multi_node_tp_pp", "multi_node_tep", "multi_node_dep" ], "features": [ "tool_calling", "reasoning" ], "opt_in_features": [], "url": "/zai-org/GLM-4.5V", "json": "/zai-org/GLM-4.5V.json" }, { "hf_id": "zai-org/GLM-4.6", "hf_org": "zai-org", "hf_repo": "GLM-4.6", "title": "GLM-4.6", "provider": "GLM (Z-AI)", "description": "GLM-4.6 MoE language model (~357B total parameters, BF16) with MTP speculative decoding, native tool calling and reasoning", "date_updated": "2026-04-17", "difficulty": "intermediate", "min_vllm_version": "0.11.0", "architecture": "moe", "parameter_count": "357B", "active_parameters": "32B", "context_length": 202752, "tasks": [ "text" ], "performance_headline": "Updated GLM-4.X series MoE model with native FP8 and BF16, MTP speculative decoding", "variants": { "default": { "precision": "bf16", "vram_minimum_gb": 857 }, "fp8": { "precision": "fp8", "vram_minimum_gb": 428 } }, "compatible_strategies": [ "single_node_tp", "single_node_tep", "single_node_dep", "multi_node_tp", "multi_node_tp_pp", "multi_node_dep", "multi_node_tep" ], "features": [ "tool_calling", "reasoning", "spec_decoding" ], "opt_in_features": [ "spec_decoding" ], "url": "/zai-org/GLM-4.6", "json": "/zai-org/GLM-4.6.json" }, { "hf_id": "zai-org/GLM-4.6V", "hf_org": "zai-org", "hf_repo": "GLM-4.6V", "title": "GLM-4.6V", "provider": "GLM (Z-AI)", "description": "GLM-4.6 vision-language MoE model — image-text-to-text with 128K context, native FP8 checkpoint, and expert parallelism", "date_updated": "2026-04-17", "difficulty": "intermediate", "min_vllm_version": "0.12.0", "architecture": "moe", "parameter_count": "107B", "active_parameters": "12B", "context_length": 131072, "tasks": [ "multimodal" ], "performance_headline": "Updated GLM-V series with 128K context length and native FP8", "variants": { "default": { "precision": "bf16", "vram_minimum_gb": 257 }, "fp8": { "precision": "fp8", "vram_minimum_gb": 128 } }, "compatible_strategies": [ "single_node_tp", "single_node_tep", "single_node_dep", "multi_node_tp", "multi_node_tp_pp", "multi_node_tep", "multi_node_dep" ], "features": [ "tool_calling", "reasoning" ], "opt_in_features": [], "url": "/zai-org/GLM-4.6V", "json": "/zai-org/GLM-4.6V.json" }, { "hf_id": "zai-org/GLM-4.7", "hf_org": "zai-org", "hf_repo": "GLM-4.7", "title": "GLM-4.7", "provider": "GLM (Z-AI)", "description": "GLM-4.7 MoE language model (~358B total parameters) with MTP speculative decoding, updated tool call parser, and reasoning support", "date_updated": "2026-04-17", "difficulty": "intermediate", "min_vllm_version": "0.11.0", "architecture": "moe", "parameter_count": "358B", "active_parameters": "32B", "context_length": 202752, "tasks": [ "text" ], "performance_headline": "Latest GLM-4.X release with updated glm47 tool call parser and MTP speculative decoding", "variants": { "default": { "precision": "bf16", "vram_minimum_gb": 859 }, "fp8": { "precision": "fp8", "vram_minimum_gb": 430 }, "nvfp4": { "precision": "nvfp4", "vram_minimum_gb": 215 } }, "compatible_strategies": [ "single_node_tp", "single_node_tep", "single_node_dep", "multi_node_tp", "multi_node_tp_pp", "multi_node_dep", "multi_node_tep" ], "features": [ "tool_calling", "reasoning", "spec_decoding" ], "opt_in_features": [ "spec_decoding" ], "url": "/zai-org/GLM-4.7", "json": "/zai-org/GLM-4.7.json" }, { "hf_id": "zai-org/GLM-5.1", "hf_org": "zai-org", "hf_repo": "GLM-5.1", "title": "GLM-5.1", "provider": "GLM (Z-AI)", "description": "GLM-5.1 refreshed version of GLM-5 — frontier-scale MoE language model (~744B total parameters) with MTP speculative decoding and thinking mode", "date_updated": "2026-04-17", "difficulty": "advanced", "min_vllm_version": "0.19.0", "architecture": "moe", "parameter_count": "744B", "active_parameters": "40B", "context_length": 202752, "tasks": [ "text" ], "performance_headline": "Refreshed GLM-5 series MoE with improved reasoning, coding, and agentic performance", "variants": { "default": { "precision": "bf16", "vram_minimum_gb": 1786 }, "fp8": { "precision": "fp8", "vram_minimum_gb": 893 } }, "compatible_strategies": [ "single_node_tp", "single_node_tep", "multi_node_tp", "multi_node_tp_pp", "multi_node_tep", "multi_node_dep" ], "features": [ "tool_calling", "reasoning", "spec_decoding" ], "opt_in_features": [ "spec_decoding" ], "url": "/zai-org/GLM-5.1", "json": "/zai-org/GLM-5.1.json" }, { "hf_id": "zai-org/GLM-5", "hf_org": "zai-org", "hf_repo": "GLM-5", "title": "GLM-5", "provider": "GLM (Z-AI)", "description": "GLM-5 frontier-scale MoE language model (~744B total parameters, 28.5T training tokens) with asynchronous RL infrastructure for reasoning, coding, and agentic tasks", "date_updated": "2026-04-17", "difficulty": "advanced", "min_vllm_version": "0.16.0", "architecture": "moe", "parameter_count": "744B", "active_parameters": "40B", "context_length": 202752, "tasks": [ "text" ], "performance_headline": "Frontier-scale MoE with 744B parameters, best-in-class open-source performance on reasoning/coding/agents", "variants": { "default": { "precision": "bf16", "vram_minimum_gb": 1786 }, "fp8": { "precision": "fp8", "vram_minimum_gb": 893 }, "nvfp4": { "precision": "nvfp4", "vram_minimum_gb": 446 } }, "compatible_strategies": [ "single_node_tp", "single_node_tep", "multi_node_tp", "multi_node_tp_pp", "multi_node_tep", "multi_node_dep" ], "features": [ "tool_calling", "reasoning", "spec_decoding" ], "opt_in_features": [ "spec_decoding" ], "url": "/zai-org/GLM-5", "json": "/zai-org/GLM-5.json" }, { "hf_id": "zai-org/GLM-ASR-Nano-2512", "hf_org": "zai-org", "hf_repo": "GLM-ASR-Nano-2512", "title": "GLM-ASR-Nano-2512", "provider": "GLM (Z-AI)", "description": "Open-source speech recognition model (~2B) with strong dialect support (Cantonese and others) and robust low-volume speech transcription", "date_updated": "2026-04-17", "difficulty": "beginner", "min_vllm_version": "0.14.1", "architecture": "dense", "parameter_count": "2.3B", "active_parameters": "1.5B", "context_length": 8192, "tasks": [ "multimodal" ], "performance_headline": "Outperforms Whisper V3 on multiple benchmarks at compact 1.5B active / 2B total size", "variants": { "default": { "precision": "bf16", "vram_minimum_gb": 11 } }, "compatible_strategies": [ "single_node_tp", "multi_node_tp", "multi_node_tp_pp" ], "features": [], "opt_in_features": [], "url": "/zai-org/GLM-ASR-Nano-2512", "json": "/zai-org/GLM-ASR-Nano-2512.json" }, { "hf_id": "zai-org/GLM-Image", "hf_org": "zai-org", "hf_repo": "GLM-Image", "title": "GLM-Image", "provider": "GLM (Z-AI)", "description": "Hybrid autoregressive + diffusion image generation model — text-to-image and image-to-image with strong text rendering and knowledge-intensive generation", "date_updated": "2026-04-17", "difficulty": "intermediate", "min_vllm_version": "0.11.0", "architecture": "dense", "parameter_count": "16B", "active_parameters": "16B", "context_length": 4096, "tasks": [ "multimodal" ], "performance_headline": "9B AR generator + 7B DiT decoder, state-of-the-art text rendering in generated images", "variants": { "default": { "precision": "bf16", "vram_minimum_gb": 38 } }, "compatible_strategies": [ "single_node_tp", "multi_node_tp", "multi_node_tp_pp" ], "features": [], "opt_in_features": [], "url": "/zai-org/GLM-Image", "json": "/zai-org/GLM-Image.json" }, { "hf_id": "zai-org/GLM-OCR", "hf_org": "zai-org", "hf_repo": "GLM-OCR", "title": "GLM-OCR", "provider": "GLM (Z-AI)", "description": "GLM-OCR image-to-text model with built-in MTP speculative decoding for high-throughput OCR serving", "date_updated": "2026-04-17", "difficulty": "beginner", "min_vllm_version": "0.12.0", "architecture": "dense", "parameter_count": "0.9B", "active_parameters": "0.9B", "context_length": 131072, "tasks": [ "multimodal" ], "performance_headline": "Multilingual end-to-end OCR VLM with MTP-accelerated decoding", "variants": { "default": { "precision": "bf16", "vram_minimum_gb": 2 } }, "compatible_strategies": [ "single_node_tp", "multi_node_tp", "multi_node_tp_pp" ], "features": [ "spec_decoding" ], "opt_in_features": [], "url": "/zai-org/GLM-OCR", "json": "/zai-org/GLM-OCR.json" }, { "hf_id": "zai-org/Glyph", "hf_org": "zai-org", "hf_repo": "Glyph", "title": "Glyph", "provider": "GLM (Z-AI)", "description": "Visual-text compression framework that renders long text into images and processes them with a reasoning VLM, scaling effective context length", "date_updated": "2026-04-17", "difficulty": "intermediate", "min_vllm_version": "0.11.0", "architecture": "dense", "parameter_count": "10B", "active_parameters": "10B", "context_length": 131072, "tasks": [ "multimodal" ], "performance_headline": "Reasoning multimodal model for visual-text compression, single-GPU deployable", "variants": { "default": { "precision": "bf16", "vram_minimum_gb": 24 } }, "compatible_strategies": [ "single_node_tp", "multi_node_tp", "multi_node_tp_pp" ], "features": [ "reasoning" ], "opt_in_features": [], "url": "/zai-org/Glyph", "json": "/zai-org/Glyph.json" } ]