AlibabaQwen3 VL 235B A22B Instruct
Qwen3 VL 235B A22B Instruct is an Alibaba open-weights model, released 23 Sept 2025. Its best configuration (thinking reasoning effort) ranks #353 of 785 on the BenchLeader Index at 51.0 ±3.8, in the upper half. The ± is the point: 309 other configurations score within that range, so they and this one cannot be told apart on quality alone — price and speed are what separate them. It scores highest in coding (58) and lowest in knowledge (41). At $0.513 per million tokens blended it is mid-priced. Output speed of 39 tokens per second puts it in the slowest quarter, with a first answer in 1.3 s. It has been measured at 2 reasoning-effort settings; this summary describes the best-scoring one, and the tabs above switch between them. Last measured 22 Oct 2026.
- Blended price
- $0.513/M
- $0.220 in · $1.39 out
- Output speed
- 39 tok/s
- measured by Artificial Analysis
- First answer
- 1.27 s
- first token 1.27 s
- Context
- 131k
- Full answer
- 14 s
- median, reasoning included
- Cost per run
- –
- Released
- 23 Sept 2025
- Overall index51
- Reasoning46
- Coding58
- Agents & tools44
- Maths57
- Knowledge41
- Instruction following57
- Human preference57
- Multimodal52
- Long context57
- Composite45
Reasoning-effort configurations
The same model behaves differently depending on how much it is allowed to think. Each row is one setting, scored only on the benchmarks that were run at that setting. “Not stated” collects results from publishers that did not say which setting they used; for a reasoning model that is usually its thinking mode, but we do not assume it. Pick a setting here or at the top of the page to see its price, speed and category scores.
| Effort | Index | Rank | Speed | First answer | Chat reply cost | Categories |
|---|---|---|---|---|---|---|
| thinkingbest | 51.0 | #353 | 39 tok/s | 1.27 s | $0.0005 | Agents & tools 44 · Coding 58 · Composite 45 · Human preference 57 · Instruction following 57 · Knowledge 41 · Long context 57 · Maths 57 · Multimodal 52 · Reasoning 46 |
| not stated | 48.7 | #417 | 39 tok/s | 1.27 s | $0.0005 | Agents & tools 40 · Coding 59 · Composite 41 · Human preference 59 · Instruction following 45 · Knowledge 38 · Long context 40 · Maths 58 · Multimodal 57 · Reasoning 47 |
How its index has moved
9 Sept 2026 to 21 Oct 2026- thinking
- not stated
The index is recomputed from scratch every day, so a line moves when a new benchmark result lands, when a publisher revises a score, or when the models it is normalised against change. Early movement usually means the score is still settling.
Benchmark results
See all LLM benchmarks. One column per reasoning effort. Rank is among every configuration of every model on that benchmark. Hover a score for the run it came from.
Reasoning
| Benchmark | thinking | not stated | Source | Trend |
|---|---|---|---|---|
| Arena Hard Prompts | 1418#154 | 1438#127 | Arena (LMArena) | |
| GPQA Diamond (AA)not in index | 77.2%#223 | 71.2%#288 | Artificial Analysis | |
| Humanity's Last Exam (AA)not in index | 11.9%#287 | 6.6%#384 | Artificial Analysis | |
| CritPt | 0.0%#315 | 0.0%#315 | Artificial Analysis |
Coding
| Benchmark | thinking | not stated | Source | Trend |
|---|---|---|---|---|
| Arena Coding | 1452#144 | 1463#125 | Arena (LMArena) |
Agents & tools
| Benchmark | thinking | not stated | Source | Trend |
|---|---|---|---|---|
| Terminal-Bench Hard | 11.4%#215 | 6.8%#249 | Artificial Analysis | |
| τ²-Bench Telecom (AA)not in index | 54.1%#188 | 35.1%#236 | Artificial Analysis |
Maths
| Benchmark | thinking | not stated | Source | Trend |
|---|---|---|---|---|
| Arena Maths | 1400#150 | 1409#138 | Arena (LMArena) |
Knowledge
| Benchmark | thinking | not stated | Source | Trend |
|---|---|---|---|---|
| AA-Omniscience | -46.5#362 | -52.7#412 | Artificial Analysis | |
| AA-Omniscience: accuracynot in index | 20.9%#290 | 20.3%#303 | Artificial Analysis | |
| AA-Omniscience: non-hallucinationnot in index | 14.8%#342 | 8.4%#460 | Artificial Analysis |
Instruction following
| Benchmark | thinking | not stated | Source | Trend |
|---|---|---|---|---|
| IFBench | 56.5%#143 | 42.6%#236 | Artificial Analysis | |
| Arena Instruction Followingnot in index | 1383#158 | 1409#124 | Arena (LMArena) |
Human preference
| Benchmark | thinking | not stated | Source | Trend |
|---|---|---|---|---|
| Arena Text | 1395#155 | 1413#136 | Arena (LMArena) | |
| Arena Creative Writingnot in index | 1338#177 | 1357#162 | Arena (LMArena) | |
| Arena Multi-turnnot in index | 1385#171 | 1425#120 | Arena (LMArena) | |
| Arena Longer Queriesnot in index | 1404#154 | 1420#131 | Arena (LMArena) |
Multimodal
| Benchmark | thinking | not stated | Source | Trend |
|---|---|---|---|---|
| Arena Vision | 1207#88 | 1247#66 | Arena (LMArena) | |
| MMMU-Pro | 68.7%#162 | 67.6%#166 | Artificial Analysis | |
| MMMU (validation)effort not stated | – | 78.7%#8 | MMMU | |
| MMMU-Pro (official)not in indexeffort not stated | – | 68.1%#11 | MMMU |
Long context
| Benchmark | thinking | not stated | Source | Trend |
|---|---|---|---|---|
| AA-LCR | 63.7%#264 | 32.7%#382 | Artificial Analysis |
Composite
| Benchmark | thinking | not stated | Source | Trend |
|---|---|---|---|---|
| AA Intelligence Index v4.3.2 | 13.4#308 | 9.9#385 | Artificial Analysis |
Where to run it
Every provider serving this model through OpenRouter, with throughput and first-token latency measured on live traffic over the last 30 minutes and each provider’s own price. Purple marks the best in each column.
| Provider | Speed | First token | Input $/M | Output $/M | Context | Quantisation |
|---|---|---|---|---|---|---|
| Venice | 44 tok/s | 1.29 s | $0.210 | $1.90 | 128k | fp8 |
| Parasail | 41 tok/s | 0.63 s | $0.210 | $1.90 | 131k | fp8 |
| DeepInfra | 13 tok/s | 2.10 s | $0.200 | $0.880 | 262k | fp8 |
Price history
The price above, per 1M tokens, every day it changed.
What a task costs
Estimates from list price, output speed and time to first answer for the best configuration. “With caching” assumes three-quarters of the input is served from the prompt cache at $0.110 per 1M. Reasoning tokens are not modelled.
| Workload | Tokens in / out | Cost | With caching | Time |
|---|---|---|---|---|
| Chat reply | 400 / 300 | $0.0005 | $0.0005 | 9.0 s |
| Summarize a 30-page report | 12,000 / 600 | $0.0035 | $0.0025 | 16.8 s |
| Code edit | 6,000 / 1,500 | $0.0034 | $0.0029 | 40.0 s |
| Agentic coding session | 60,000 / 4,000 | $0.019 | $0.014 | 1.7 min |
| Structured extraction | 2,000 / 200 | $0.0007 | $0.0006 | 6.4 s |
More from Alibaba
See also
Questions people ask
- What are Qwen3 VL 235B A22B Instruct's benchmark scores?
- As of 25 Oct 2026, Qwen3 VL 235B A22B Instruct (thinking effort) scores 51.0 ±3.8 on the BenchLeader Index, strongest in coding (58) and weakest in knowledge (41). That rests on 12 independent benchmark results, and every score on this page links to its publisher.
- How does Qwen3 VL 235B A22B Instruct rank?
- Qwen3 VL 235B A22B Instruct (thinking effort) is #353 of 785 ranked configurations on 25 Oct 2026. 309 other configurations score within its margin, so the exact position matters less than the band. The leader is Claude Opus 5.5 (max effort) at 72.6.
- How much does Qwen3 VL 235B A22B Instruct cost per 1M tokens?
- As of 25 Oct 2026, Qwen3 VL 235B A22B Instruct costs $0.220 per 1M input tokens and $1.39 per 1M output tokens, $0.513 blended (three input tokens per output token). The list price last changed on 13 Oct 2026.
- When was Qwen3 VL 235B A22B Instruct released?
- Alibaba released Qwen3 VL 235B A22B Instruct on 23 Sept 2025. BenchLeader has results for it at 2 reasoning-effort settings.
Data as of 25 Oct 2026. Compare these configurations.
Cite as: BenchLeader, “Qwen3 VL 235B A22B Instruct: benchmarks, pricing, speed and rank”, data as of 25 Oct 2026, https://www.benchleader.com/models/qwen3-vl-235b-a22b.