Which Chinese open-weight model families are still worth watching now?

Qwen and DeepSeek remain the two families most builders should verify first, because they combine capability, public release surfaces, and comparatively clear licensing paths. GLM, InternLM, CogVideoX, Hunyuan-related releases, and other Chinese open-weight projects can still matter, but they should be checked against the current official release page or model card before you treat them as active default options.

Is Qwen3 Apache 2.0 licensed for commercial use?

Qwen is still the clearest current reference point for a permissive commercial open-weight path, but the safest practice is to verify the specific current release page and LICENSE file for the exact family or branch you are deploying. Use the official Qwen release surface rather than a static third-party summary.

What open source Chinese LLMs can I use for commercial fine-tuning?

Qwen remains the safest default route for commercial fine-tuning because of license clarity and tooling support. DeepSeek-distilled routes remain relevant for reasoning-heavy use cases. Other Chinese families can still matter, but the right move is to verify the current release surface and current license before committing to a fine-tuning path.

How does DeepSeek-R1 MIT license compare to Qwen Apache 2.0?

MIT is marginally more permissive than Apache 2.0 in practice. MIT requires only copyright notice and license text preservation. Apache 2.0 additionally requires a NOTICE file attribution if present, and has patent grant clauses (which protect users from patent claims). For most commercial deployments, both licenses are equivalent. The key practical difference: Apache 2.0's explicit patent grant is better protection for enterprise legal teams. MIT's shorter text is easier to clear in legal review. Both allow commercial use, fine-tuning, hosting as a service, and redistribution of derivatives without notifying the original author.

Are there any restrictions on using Chinese open-source AI models globally?

For Qwen3 (Apache 2.0) and DeepSeek-V3/R1 (MIT): no geographic restrictions, no export controls imposed by the license itself. US export control law applies to the downstream application (e.g., if you deploy to sanctioned entities), but the model licenses themselves impose no country restrictions. Baichuan2 has an MAU cap (custom license kicks in above 100M monthly active users). ERNIE and Kimi are proprietary API-only and not available as weights. Always verify the LICENSE file in each HuggingFace repo before commercial deployment — license terms can differ between model size variants.

Where can I download open source Chinese AI models?

Primary download surfaces: Qwen3 at huggingface.co/Qwen (all sizes, GGUF and FP16); DeepSeek-V3 and R1 at huggingface.co/deepseek-ai; MiniMax-Text-01 at huggingface.co/MiniMaxAI; GLM-4-9B at huggingface.co/THUDM; InternLM2 at huggingface.co/internlm. China-domestic mirror: modelscope.cn hosts all major Chinese open-source models with faster download speeds from within China. Ollama library has quantized GGUF versions of Qwen3-0.6B through Qwen3-14B for local inference.

China Open Source AI Models (Apache 2.0 / MIT, 2026 List)

Thesis

The key open-weight question in 2026 is not "which single Chinese model is best?" but "which families are still active, deployable, and easy to verify." Qwen remains the clearest public Apache-style route to watch. DeepSeek remains the clearest current hosted-plus-open-weight route to watch. Other Chinese open-weight families still matter, but they should be validated against their current release surface before being treated as default options in a builder stack.

Decision in 20 seconds

Use case	Best open-source option	License
Commercial fine-tuning (general purpose)	Qwen3-7B or Qwen3-14B	Apache 2.0
Fine-tuning for reasoning / STEM	DeepSeek-R1-Distill-Qwen-7B	MIT
On-prem inference, single A100 40GB	Qwen3-30B-A3B MoE (full precision) or Qwen3-32B (FP8)	Apache 2.0
Local inference, M-series Mac	Qwen3-8B-Q4_K_M (Ollama) or DeepSeek-R1-7B (Ollama)	Apache 2.0 / MIT
Bilingual Chinese-English fine-tuning	GLM-4-9B or InternLM2-7B	Apache 2.0

Open source Chinese AI models — full list (2026)

Family	Lab	What to verify now	License posture	Primary download surface	Commercial-use note
Qwen family	Alibaba	Current active branch, current LICENSE, and which sizes are still being foregrounded on the official release surface	Use current official release page as source of truth	huggingface.co/Qwen / QwenLM GitHub	Best current open-weight default for many builder workflows once the current LICENSE is confirmed
DeepSeek family	DeepSeek	Current hosted-vs-open-weight branches, current license surface, and which variants still matter for builders	Verify at current official release or model-card surface	huggingface.co/deepseek-ai	Important when reasoning, coding, and hosted/open-weight crossover all matter
MiniMax family	MiniMax	Which public weights, multimodal branches, and current platform-facing routes are still active	Verify branch-specific license and hosting path	huggingface.co/MiniMaxAI / official platform	Relevant when multimodal and long-context product surfaces matter
GLM / THUDM family	Z.AI / THUDM	Which branches are open-weight, which are hosted, and which current docs are still active	Mixed depending on branch	huggingface.co/THUDM / docs.z.ai	Relevant when bilingual, multimodal, or agent/coding routes matter
InternLM family	Shanghai AI Lab	Whether the family remains relevant for your use case and whether the current release surface is still active	Apache-style open route	huggingface.co/internlm	More niche than Qwen/DeepSeek, but still useful for selective workloads
Baichuan family	Baichuan AI	Whether the current custom-license branch still fits your commercial scale	Custom license	huggingface.co/baichuan-inc	Verify commercial restrictions directly on the current license before use
CogVideoX and related video families	Z.AI / THUDM	Whether the current video branch is still the right open-weight route for your workflow	Check current release surface	huggingface.co/THUDM	Useful when video-generation capability is the real builder question

License comparison — what matters for your product

License type	Models using it	Can fine-tune commercially?	Can host as a service (SaaS)?	Key requirement	Enterprise legal notes
Apache 2.0	Qwen3, GLM-4-9B, InternLM2, CogVideoX	Yes, without restriction	Yes, without restriction	Preserve copyright notice + NOTICE file if present	Patent grant clause protects licensee from patent claims by licensor — preferred by many enterprise legal teams over MIT
MIT	DeepSeek-V3, DeepSeek-R1, MiniMax-Text-01	Yes, without restriction	Yes, without restriction	Preserve copyright notice + MIT license text in distributions	Shorter and simpler than Apache 2.0; no explicit patent grant (slight theoretical risk); faster legal review cycle
Custom (Baichuan)	Baichuan2-7B/13B	Yes, below 100M MAU	Yes, below 100M MAU	Contact Baichuan AI for commercial license if MAU exceeds 100M	Practical blocker for large consumer apps; fine for enterprise internal tools or smaller SaaS products

Why license routing matters for open-source Chinese models

The license landscape for Chinese open-source AI improved dramatically between 2023 and 2026. In 2023, most major Chinese models used custom licenses with vague commercial restrictions — LLaMA-style constraints that made enterprise deployment legally uncertain. The shift to Apache 2.0 (Qwen family from Qwen1.5 onwards) and MIT (DeepSeek from V2 onwards) resolved this entirely for the two strongest model families. The practical result: a team building a commercial product in 2026 has two fully permissive routes with state-of-the-art capability, neither requiring negotiation or legal review beyond standard OSS process. The main remaining exception is Baichuan2 — useful at smaller scales but the MAU cap makes it unsuitable for growth-stage consumer products.

Infrastructure is the more meaningful constraint now. Qwen3-235B-A22B at full precision requires 4×H100 for inference. The MoE architecture (only 22B parameters active per forward pass) makes it much more efficient than a dense 235B model, but it still requires significant hardware. For teams without that infrastructure: Qwen3-14B on a single A100 40GB delivers strong capability; Qwen3-30B-A3B MoE is the best quality-per-GPU-cost option at single-node scale.

FAQ

What Chinese AI models are open source and commercially usable?: Qwen3 (all sizes, Apache 2.0, Alibaba), DeepSeek-V3 and R1 (MIT), MiniMax-Text-01 (MIT), GLM-4-9B (Apache 2.0), InternLM2 (Apache 2.0). Baichuan2 is commercial below 100M MAU. ERNIE 4.0 and Kimi are proprietary API-only.
Is Qwen3 Apache 2.0 licensed for commercial use?: Yes — all Qwen3 sizes (0.6B through 235B-A22B MoE) are Apache 2.0 with no MAU cap, no revenue threshold, and no notification requirement. Attribution (preserving the license notice) is the only obligation. License text at github.com/QwenLM/Qwen3.
What open source Chinese LLMs can I use for commercial fine-tuning?: Qwen3-7B/14B (Apache 2.0) for general tasks; DeepSeek-R1-Distill-Qwen-7B (MIT) for reasoning-heavy tasks; InternLM2-7B (Apache 2.0) for Chinese scientific/technical text; GLM-4-9B (Apache 2.0) for bilingual fine-tuning. All support LLaMA-Factory, Axolotl, or Unsloth.
How does DeepSeek-R1 MIT license compare to Qwen Apache 2.0?: Both fully permit commercial use, fine-tuning, and hosting as a service. MIT is simpler (faster legal review). Apache 2.0 adds an explicit patent grant (protects licensee from patent claims). In practice: pick the model with better capability for your task — the license difference is rarely the deciding factor.
Are there any restrictions on using Chinese open-source AI models globally?: Qwen3 (Apache 2.0) and DeepSeek (MIT) impose no geographic restrictions. US export control law governs your downstream application but not the model weights themselves. Baichuan2 has an MAU cap in its custom license. Always verify the specific LICENSE file in each HuggingFace repo variant.
Where can I download open source Chinese AI models?: Qwen3: huggingface.co/Qwen. DeepSeek: huggingface.co/deepseek-ai. MiniMax-Text-01: huggingface.co/MiniMaxAI. GLM-4-9B: huggingface.co/THUDM. InternLM2: huggingface.co/internlm. China mirror: modelscope.cn. Quantized local inference: Ollama library (Qwen3 0.6B–14B, DeepSeek-R1 7B).

Companion pages in this cluster

If your question is about…	Go to	What's there
Full benchmark and API comparison of China AI models	China AI Foundation Models	Qwen3, DeepSeek, Kimi, GLM, ERNIE — benchmarks, API access, use-case routing
Recent model releases and capability milestones	Model Release Tracker	Release timeline with benchmark data and verification sources
Weekly digest of what changed in China AI	China AI Updates	Weekly signal digest — model releases, funding, policy, curated for builders
Which China AI companies to watch (including open-source labs)	Best China AI Companies	15-company shortlist: foundation model labs, workflow infra, physical AI
China AI startup funding and investment rounds	China AI Startup Funding Tracker	Funding tracking logic by company category and current market surface

Quotable summary: In 2026, the license question for Chinese open-source AI is largely settled: Qwen3 (Apache 2.0) and DeepSeek (MIT) provide fully permissive commercial use for the two most capable open-weight model families in the world. The remaining decision is infrastructure — which GPU tier you have — and use-case fit: Qwen3 for multilingual and general fine-tuning, DeepSeek for code and reasoning. Everything else (Baichuan, GLM, InternLM) fills narrower niches with the same open-license principle.