Chinese open-source AI models (which repos deserve weekly checks)

Decision in 20 seconds

If you want to track Chinese open-source AI models well, ground the routine in repo movement, model cards, and usable release paths. RadarAI is the first routing layer, but the actual decision comes from whether the repo, weights, docs, and license create a realistic path to testing.

Key points

QwenLM released Qwen-Image-2.1 (7B), unifying image generation and editing; integrated into ComfyUI and vLLM-Omni.
Qwen3.8-LiveTranslate reduced real-time speech translation latency to 2.3 seconds using a Thinker–Talker architecture.
DeepSeek and THUDM maintain active GitHub repositories, but no recent updates were observed in the evidence for September 2026.

What changed recently

Qwen-Image-2.1 launched on 2026-09-21, supporting unified image tasks.
Qwen3.8-LiveTranslate launched on 2026-09-20, targeting low-latency speech translation.

Explanation

This page is maintained as an evergreen knowledge page. It prioritizes clarity, trade-offs, and verifiable sources.

Tools / Examples

QwenLM/Qwen2-VL on GitHub: supports vision-language tasks with clear versioning and inference examples.
DeepSeek-AI/deepseek-moe on GitHub: a sparse MoE model with documented quantization and fine-tuning guides.

Evidence timeline

Alibaba's Tongyi Qianwen Open-Sources Qwen-Image-2.1, a 7B Model for Unified Image Generation and Editing · 0921-669

2026-09-21

Alibaba's Tongyi Qianwen has open-sourced Qwen-Image-2.1, unifying image generation and editing in a single 7B parameter model, quickly gaining ecosystem support from ComfyUI, vLLM-Omni, and others [7][16][17]; meanwhile

Alibaba's Qwen3.8 Cuts Real-Time Translation Latency to 2.3 Seconds; Google's Gemini 4 Pro Sees Major Performance Gains · 0920-667

2026-09-20

The AI field saw a flurry of updates this week: Alibaba released the Qwen3.8-LiveTranslate real-time speech translation model, reducing latency to 2.3 seconds through a Thinker–Talker architecture and camera-frame-assist

Anthropic Builds Its Own Wet Lab, Google Gemini Moves Into Real Companies, AI Trust Crisis Deepens · 0920-666

2026-09-20

The AI industry is simultaneously experiencing two main threads: "capability leaps" and a "trust crisis." On one hand, Anthropic has been revealed to be building its own wet lab to accelerate its AI pharmaceutical closed

Sources

FAQ

Which Chinese open-source AI model repos should I check weekly?

Prioritize QwenLM (QwenLM/Qwen2-*), DeepSeek (deepseek-ai/*), and THUDM (THUDM/*) on GitHub — they are consistently updated, though recent evidence only confirms QwenLM activity in late September 2026.

Is there evidence of THUDM or DeepSeek releasing new models recently?

No — the evidence does not include verified releases or major updates from THUDM or DeepSeek in the cited timeframe. Their GitHub repos remain valid sources, but recent change activity is unconfirmed here.

Search angles this page supports

Chinese open-source AI models open source Qwen DeepSeek THUDM GitHub

Last updated: 2026-09-22 · Policy: Editorial standards · Methodology