// Official vendor repos
魔搭 ModelScope
7 tracked official repos · 4 releases in the last 30 days.
01Latest releases
-
modelscope/evalscope v1.14.0 Minor New model
决策评测：新增 JevBench、OpenDecisions、STS-B 和自定义 JSONL 数据集 `general_decision`，覆盖 Choice、Noul 和 Score 任务。
-
modelscope/ms-swift v4.6.0 Minor New model
JEV Support (typed-decision / System-1)
-
modelscope/evalscope v1.13.0 Minor
多轮对话评测：新增 MT-Bench，支持两轮对话和可配置的 Judge，遵循官方评分流程。
-
modelscope/modelscope v1.40.2 Patch
Add Intel XPU device support** — Added support for Intel XPU (including iGPU and discrete GPUs such as Arc, Flex, and Max series) as a target device for model inference and training. This enables running ModelScope model
02All repos
| Repo | Category | ★ | Latest | Licence |
|---|---|---|---|---|
| FunASROpen-source speech recognition toolkit for training, inference, streaming ASR, VAD, punctuation, speaker diarization pipelines, and OpenAI-compatible/MCP serving. | Speech & audio | 20.8k | v1.4.162026-09-18 | MIT |
| ms-swiftUse PEFT or Full-parameter to CPT/SFT/DPO/GRPO 600+ LLMs (Qwen3.8, DeepSeek-V4, GLM-5.1, InternLM3, Llama4, ...) and 300+ MLLMs (Qwen3-VL, Qwen3-Omni, InternVL3.5, Ovis2.5, GLM5… | Training & fine-tuning | 15.9k | v4.6.02026-10-13 | Apache-2.0 |
| DiffSynth-StudioEnjoy the magic of Diffusion models! | Image, video & 3D generation | 13.3k | v1.1.92025-11-18 | Apache-2.0 |
| facechainFaceChain is a deep-learning toolchain for generating your Digital-Twin. | Image, video & 3D generation | 9,508 | v2.0.02023-12-20 | Apache-2.0 |
| modelscopeModelScope: bring the notion of Model-as-a-Service to life. | Model weights & research code | 9,181 | v1.40.22026-10-09 | Apache-2.0 |
| ms-agentMS-Agent: a lightweight framework to empower agentic execution of complex tasks | Deep research & AI search | 4,417 | v1.6.02026-03-23 | Apache-2.0 |
| evalscopeA streamlined and customizable framework for efficient large model (LLM, VLM, AIGC) evaluation and performance benchmarking. | RAG & knowledge bases | 3,572 | v1.14.02026-10-20 | Apache-2.0 |