Qwen3-VL-235B-A22B-Thinking
- In Input modalities: TextImageVideo Out Output modalities: Text
- Reasoning
- Tool use
- Structured output
- Open weights
Prices
3 offerings · prices as listedqwen3-vl-235b-a22b-thinking| Conditions/ 1M tokens | Input | Output |
|---|---|---|
| Standard no longer listed since Oct 11, 2026 | $0.40 | $4.00 |
Prices from Alibaba Cloud Model Studio pricing (opens in a new tab) (checked Oct 11, 2026)
qwen3-vl-235b-a22b-thinking| Conditions/ 1M tokens | Input | Output |
|---|---|---|
| Standard no longer listed since Oct 11, 2026 | $0.287 | $2.867 |
Prices from Alibaba Cloud Model Studio pricing (opens in a new tab) (checked Oct 11, 2026)
qwen3-vl-235b-a22b-thinking| Conditions/ 1M tokens | Input | Output |
|---|---|---|
| Standard no longer listed since Oct 11, 2026 | ¥2.00 | ¥20.00 |
Prices from 阿里云百炼：模型价格 (opens in a new tab) (checked Oct 6, 2026)
Standard: the price a request with no conditions pays (the headline price elsewhere on the site).
Details
- Max thinking budget
- 81.92K
- Knowledge cutoff
- Mar 31, 2025
- Family
- qwen-vl-thinking
- Model ID
Not stated on Alibaba Qwen’s pages: max input, reasoning efforts, reasoning can be off, built-in tools, licence, aliases.
Run locally
Weights: Qwen/Qwen3-VL-235B-A22B-Thinking- Parameters 235.7B
- Active 22.8B (estimated) · 128 experts, 8 per token
- Native precision BF16
- Layers 94
- KV cache per token 188 KB
- Config positions 256K (config.json max_position_embeddings)
- API context 128K (maker’s API)
- Needs
- ≈ 123.9 GB (weights 116.6 + KV cache 1.5 + overhead 5.8)
- Usable
- 21.6 GB (90% of its memory)
- Try
- IQ2_XXS runs (needs system memory), but below 4 bits quality drops noticeably
| Precision | Weights | Needs at 8K | On your hardware | Longest context |
|---|---|---|---|---|
| BF16 (native)maker’s file | 439.0 GB | ≈ 462.4 | ✕Doesn’t fit | — |
| Q8_0maker’s file | 232.8 GB | ≈ 245.9 | ✕Doesn’t fit | — |
| FP8maker’s file | 221.3 GB | ≈ 233.8 | ✕Doesn’t fit | — |
| Q6_K | 180.3 GB | ≈ 190.7 | ✕Doesn’t fit | — |
| Q5_K_M | 156.1 GB | ≈ 165.4 | ✕Doesn’t fit | — |
| Q5_0 | 150.9 GB | ≈ 159.9 | ✕Doesn’t fit | — |
| Q4_K_M | 133.3 GB | ≈ 141.5 | ✕Doesn’t fit | — |
| Q4_0 | 123.5 GB | ≈ 131.1 | ✕Doesn’t fit | — |
| IQ4_XS | 116.6 GB | ≈ 123.9 | ✕Doesn’t fit | — |
| Q3_K_Mquality drops | 107.3 GB | ≈ 114.1 | ✕Doesn’t fit | — |
| IQ3_XXSquality drops | 84.0 GB | ≈ 89.6 | ✕Doesn’t fit | — |
| Q2_Kquality drops | 82.3 GB | ≈ 87.9 | ✕Doesn’t fit | — |
| IQ2_XXSquality drops | 56.5 GB | ≈ 60.8 | ○Needs system memory 12–17 tok/s | — |
Click a row to choose a precision. “Maker’s file” sizes are the maker’s own files; others are parameters × bits per weight.
Compared with other hardware
Each at the highest precision it runs. Click one for details.
Smallest that runs it (Q4_K_M, 8K): GPU NVIDIA B200 (HGX B200); unified memory Apple M5 Ultra (64-core GPU) · 256 GB.
Run commands
The maker publishes no files at this precision. Choose its native precision or one of its own builds, or quantise it yourself with llama.cpp.
Memory to fine-tune
Batch 1, 2,048-token sequences, gradient checkpointing; LoRA and QLoRA train rank-16 adapters on the attention projections.
Estimates; facts from the maker’s Hugging Face repositories, as of 2026-11-06. How it’s estimated
Cited scores
From third-party leaderboards, not measured by us- LMArena Text 1,395
#103 of 163 listed models · CI 1,388–1,402 · 7,862 votes
- LMArena Text · Chinese 1,423
#104 of 161 listed models · CI 1,395–1,451 · 443 votes
- LMArena Text · Coding 1,452
#94 of 163 listed models · CI 1,437–1,466 · 1,638 votes
- LMArena Text · Creative writing 1,338
#112 of 163 listed models · CI 1,320–1,356 · 1,021 votes
- LMArena (Arena), leaderboard-dataset on Hugging Face, licensed CC BY 4.0 · License · as of Oct 19 – 28, 2026
- changes we made
Changes
All changes →No changes recorded yet.
Sources
- Alibaba Cloud Model Studio pricing (opens in a new tab) alibabacloud.com · checked Oct 11, 2026
- 阿里云百炼：模型价格in Chinese (opens in a new tab) help.aliyun.com · checked Oct 6, 2026
- Alibaba Cloud Model Studio: newly released models (opens in a new tab) alibabacloud.com · checked Nov 2, 2026
- qwen3-vl-235b-a22b-thinking (opens in a new tab) docs.modelstudio.console.alibabacloud.com · checked Oct 7, 2026
- \N (opens in a new tab) developer.aliyun.com · checked Oct 30, 2026
- models.dev: alibaba/qwen3-vl-235b-a22b-thinking (community data, MIT) (opens in a new tab) github.com · checked Sep 30, 2026
- qwen3-vl-235b-a22b-thinking - Alibaba Cloud Model Studio (opens in a new tab) docs.modelstudio.console.alibabacloud.com · checked Oct 10, 2026
- Qwen/Qwen3-VL-235B-A22B-Thinking model card (opens in a new tab) huggingface.co · checked Oct 31, 2026
- Qwen/Qwen3-VL-235B-A22B-Thinking config.json (opens in a new tab) huggingface.co · checked Oct 31, 2026
- Qwen/Qwen3-VL-235B-A22B-Thinking files (opens in a new tab) huggingface.co · checked Oct 31, 2026
- Qwen/Qwen3-VL-235B-A22B-Thinking-FP8 files (opens in a new tab) huggingface.co · checked Nov 1, 2026
- Qwen/Qwen3-VL-235B-A22B-Thinking-GGUF files (opens in a new tab) huggingface.co · checked Nov 2, 2026
Alibaba Qwen’s official policies: data use, privacy, terms →
Some values on this page are still waiting to be checked against official pages: a filled dot is verified, a hollow one is not yet (with or without a source on file). How we verify data
Alibaba Qwen · China · qwen.ai
Related models
Snapshots, the same family and other models from this provider