Qwen3.5-35B-A3B
AlibabaOpen weightsReasoningVerifiedReleased Mar 4, 2026Input: text, image, video
Overall score
45.8
Rank 120 of 222 · 34.8–56.8
Context window
262K
262,000 tokens
Price per 1M tokens
$0.14 in · $1.00 out
Source: OpenRouter
Blended price
$0.35
3:1 input:output tokens
Output speed
—
Not measured yet
Category scores
Scores per capability area with the model's rank among ranked models in that category.
- 1Instruction following#3388.8
- 2Multilingual#1183.8
- 3Knowledge#10869.3
- 4Multimodal#4364.6
- 5Reasoning64.4
- 6Coding52.6
- 7Agentic & tool use#8617.4
Benchmark results (20)
As reported in system cards, papers and public evaluations. Position compares this model with every other model reporting the same benchmark.
Multimodal
| MMMU-Pro | 75.1 | #33 of 48 |
| CharXiv | 77.5 | #33 of 42 |
| ExtractBench (LlamaIndex)(lower is better) | 88.00 | #13 of 24 |
| MMMU | 81.4 | #10 of 21 |
| MathVision | 83.9 | #16 of 20 |
| V* | 92.7 | #8 of 12 |
| MMVU | 72.3 | #7 of 7 |
Agentic & tool use
| Gert Labs | 28.96 | #51 of 52 |
| BrowseComp | 61.0 | #39 of 50 |
| Terminal-Bench 2.0 | 40.5 | #38 of 39 |
| OSWorld-Verified | 54.5 | #33 of 35 |
Coding
| SWE-bench Verified | 69.2 | #66 of 98 |
| LiveCodeBench v6 | 74.6 | #29 of 46 |
| SWE-Rebench | 53.7 | #11 of 13 |
Knowledge
| GPQA | 84.2 | #51 of 117 |
| MMLU-Pro | 85.3 | #17 of 67 |
| SuperGPQA | 63.4 | #16 of 20 |
Reasoning
| LongBench v2 | 59.0 | #15 of 17 |
Multilingual
| MMLU-ProX | 81.0 | #11 of 16 |
Instruction following
| IFEval | 91.9 | #16 of 53 |
Pricing by source
USD per 1M tokens.
| Source | Input | Output | Cached in |
|---|---|---|---|
| List price (OpenRouter) | $0.14 | $1.00 | — |
| OpenRouter qwen/qwen3.5-35b-a3b | $0.14 | $1.00 | $0.050 |
Open-weight model. Self-hosted or third-party hosted costs vary by provider and infrastructure.
Price history
- Oct 17, 2026: $0.15 / $1.00
- Oct 20, 2026: $0.14 / $1.00
Arena ratings
Human-preference ratings from LMArena. Best-rated variant shown.
| Text | 1395 | #161 · 30.5K votes |
| WebDev (coding) | 1254 | #135 · 1.6K votes |
Open weights
From the Hugging Face repository.
- Parameters
- 36B
- Licence
- apache-2.0
- Downloads (30 days)
- 1.4M
- Repository
- Qwen/Qwen3.5-35B-A3B ↗
Speed
Measured by us through OpenRouter (median of trials).
Not measured yet.
Related models
Same family
- Qwen3.5 Flash42.8
Similar score
- Grok 4.1 Fast (Reasoning)45.9 · —
- Gemma 4 26B A4B45.6 · $0.12
- GPT-5 (medium)46.0 · —
- o345.4 · $3.50
- o3-pro45.4 · $35.00