Model Tracker
Qwen3-235B-A22B
Qwen3-235B-A22B is an AI model made by Alibaba Tongyi. Alibaba Tongyi released it on 29 April 2025. It is generally available. It takes text and returns text.
In inglese
Qwen3-235B-A22B is an AI model made by Alibaba Tongyi. Alibaba Tongyi released it on 29 April 2025. It is generally available. It takes text and returns text. Its weights are open. Its licence is Apache-2.0.
- Released
- 29 April 2025 — qwenlm.github.io, read 16 October 2026
- Status
- Generally available — huggingface.co, read 16 October 2026
- Takes
- Text — huggingface.co, read 16 October 2026
- Returns
- Text — huggingface.co, read 16 October 2026
- Open weights
- Yes — qwenlm.github.io, read 16 October 2026
- Licence
- Apache-2.0 — huggingface.co, read 16 October 2026
- Context window
- Not confirmed: two readings disagree: 32768 and 128000
- Longest answer
- Not confirmed: The model card gives a generation example configured with 32768 new tokens and recommends that output length for most queries, but does not identify it as the standard maximum output limit.
- Knowledge cutoff
- Not confirmed: The maker's model card, announcement, and documentation reviewed do not state a knowledge-cutoff date for this model.
- API price, input
- Not confirmed: The current standard Model Studio pricing page, last updated October 15, 2026, does not list this model's standard API input price.
- API price, output
- Not confirmed: The current standard Model Studio pricing page, last updated October 15, 2026, does not list this model's standard API output price; the maker's published pricing also distinguishes thinking and non-thinking output rates.
- Where
- API, Consumer app, Downloadable weights, Cloud marketplaces — qwenlm.github.io, read 16 October 2026
- Parameters
- 235,000,000,000 — huggingface.co, read 16 October 2026
- Model card
- https://huggingface.co/Qwen/Qwen3-235B-A22B — huggingface.co, read 16 October 2026
Benchmarks
Measured by others, not by the maker: each score with the name the measurer gives the model and the setting it ran, its day and its licence.
Arena
- Text Arena, qwen3-235b-a22b: 1,375 (1,370–1,380, 25,948 votes), published 15 October 2026
Arena’s leaderboard dataset (lmarena-ai/leaderboard-dataset on Hugging Face), CC BY 4.0: ratings from people’s votes between two anonymous models, as Arena publishes them. Data · CC BY 4.0
Epoch AI
- Epoch Capabilities Index, Qwen3-235B-A22B: 139.4 (135.2–140.8), as of 19 October 2026
- GPQA Diamond, qwen3-235b-a22b: 70.7% (± 2.7), run 3 June 2025
Epoch AI, ‘Capabilities & benchmarking’, epoch.ai/benchmarks, CC BY 4.0: Epoch AI’s own runs and its Capabilities Index. Data · CC BY 4.0
Questions
When was Qwen3-235B-A22B released?
Alibaba Tongyi released Qwen3-235B-A22B on 29 April 2025.
Is Qwen3-235B-A22B still available?
Yes: it is generally available, as Alibaba Tongyi states it.
Are the weights of Qwen3-235B-A22B open?
Yes: Alibaba Tongyi publishes the weights of Qwen3-235B-A22B for download.
What is the licence of Qwen3-235B-A22B?
The licence of Qwen3-235B-A22B is Apache-2.0, as Alibaba Tongyi names it.
What does Qwen3-235B-A22B take and return?
Qwen3-235B-A22B takes text and returns text.
How many parameters does Qwen3-235B-A22B have?
Qwen3-235B-A22B has 235,000,000,000 parameters, as Alibaba Tongyi publishes it.
More from Alibaba Tongyi
- Qwen-1.8B
- Qwen-14B
- Qwen-72B
- Qwen-7B
- Qwen-Audio
- Qwen-Image
- Qwen-VL
- Qwen-VL-Max
- Qwen-VL-Plus
- Qwen1.5-MoE-A2.7B
- Qwen2-Audio
- Qwen2-VL-72B
- Qwen2.5-Omni-3B
- Qwen2.5-Omni-7B
- Qwen2.5-VL-3B
- Qwen2.5-VL-7B
- Qwen3-0.6B
- Qwen3-1.7B
- Qwen3-14B
- Qwen3-30B-A3B
- Qwen3-32B
- Qwen3-4B
- Qwen3-8B
- Qwen3-ASR-0.6B
Sources
- Qwen3: Think Deeper, Act Faster (read 16 October 2026)
- GitHub - QwenLM/Qwen3: Qwen3 is the large language model series developed by Qwen team, Alibaba Cloud. (read 16 October 2026)
- Qwen/Qwen3-235B-A22B (read 16 October 2026)
- Qwen3 - a Qwen Collection (read 16 October 2026)
- Platform For AI:Deploy, fine-tune, and evaluate Qwen3 models in PAI (read 16 October 2026)
- Alibaba Cloud Model Studio:Model inference pricing (read 16 October 2026)