AlibabaQwen3 235B A22B 2507
As of 29 Oct 2026, Qwen3 235B A22B 2507 by Alibaba scores 51.9 on the BenchLeader Index (581 ranked configurations), with its best results in long context (60) and weakest in agents & tools (39). It costs $0.385 per 1M tokens blended ($0.210 in / $0.840 out), and streams about 110 tokens/s (Artificial Analysis). Released 22 Jul 2025; 15 independent results so far.
Key facts (as of 29 Oct 2026)
- Qwen3 235B A22B 2507 costs $0.385 per 1M tokens blended ($0.210 in / $0.840 out) (list price, 29 Oct 2026).
- Qwen3 235B A22B 2507 streams about 110 tokens/s (Artificial Analysis, 29 Oct 2026).
- Qwen3 235B A22B 2507 was released on 22 Jul 2025 (Alibaba).
In detail. Qwen3 235B A22B 2507 is an Alibaba open-weights reasoning model, released 22 Jul 2025. Its best configuration (thinking reasoning effort) ranks #286 of 581 on the BenchLeader Index at 51.9 ±4.0, in the upper half. The ± is the point: 266 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 long context (60) and lowest in agents & tools (39). At $0.368 per million tokens blended it is cheaper than most ranked models. Output speed of 107 tokens per second puts it faster than most, with a first answer in 19.4 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
- $0.368/M
- $0.210 in · $0.840 out
- Output speed
- 107 tok/s
- measured by Artificial Analysis
- First answer
- 19 s
- first token 1.23 s
- Context
- 262k
- Full answer
- 24 s
- median, reasoning included
- Cost per run
- –
- Released
- 22 Jul 2025
- Overall index52
- Reasoning58
- Coding48
- Agents & tools39
- Maths58
- Knowledge47
- Instruction following53
- Human preference57
- Long context60
Versions
Alibaba has shipped 2 models under this name. Each is ranked on its own results; a newer version often has fewer results so far, which holds its index nearer the average until more arrive.
| Model | Released | Index | Rank |
|---|---|---|---|
| Qwen3 235B A22B 2507thinkingthis page | 22 Jul 2025 | 51.9 | #286 |
| Qwen3 235B A22B | 28 Apr 2025 | 50.0 | #330 |
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 | 51.9 | #286 | 107 tok/s | 19 s | $0.0003 | Agents & tools 39 · Coding 48 · Human preference 57 · Instruction following 53 · Knowledge 47 · Long context 60 · Maths 58 · Reasoning 58 |
| not stated | 50.9 | #313 | 47 tok/s | 1.23 s | $0.0003 | Agents & tools 57 · Coding 51 · Human preference 60 · Instruction following 54 · Knowledge 38 · Long context 41 · Maths 69 · Reasoning 38 |
How its index has moved
8 Sept 2026 to 29 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 |
|---|---|---|---|---|
| GPQA Diamond | 80.0%#127 | – | Epoch AI Benchmarking Hub | |
| Arena Hard Prompts | 1418#158 | 1447#114 | Arena (LMArena) | |
| GPQA Diamond (AA)not in index | 79.0%#206 | 75.3%#251 | Artificial Analysis | |
| Humanity's Last Exam (AA)not in index | 15.9%#258 | 11.1%#309 | Artificial Analysis | |
| GPQA Diamond (HELM)not in indexeffort not stated | – | 72.6%#7 | HELM Capabilities | |
| ARC-AGI-1effort not stated | – | 11.0%#235 | ARC Prize | |
| ARC-AGI-2effort not stated | – | 1.3%#219 | ARC Prize | |
| CritPtnot in index | 0.0%#328 | 0.0%#328 | Artificial Analysis | |
| Mystery Game Puzzlesnot in index | 9.0%#116 | – | Epoch AI Benchmarking Hub | |
| LMCAnot in index | 29.3%#140 | 23.5%#154 | Epoch AI Benchmarking Hub | |
| DTBenchnot in index | 80.3%#131 | 78.4%#140 | Epoch AI Benchmarking Hub | |
| ForecastBenchnot in index | 62.2%#59 | – | ForecastBench |
Coding
| Benchmark | thinking | not stated | Source | Trend |
|---|---|---|---|---|
| SciCode | 41.4%#163 | – | SciCode | |
| WeirdML | 41.0%#98 | 38.7%#111 | WeirdML | |
| Arena Coding | 1442#157 | 1472#117 | Arena (LMArena) | |
| SciCode (AA)not in index | 41.4%#175 | – | Artificial Analysis | |
| Terminal-Bench 4.0 (AA)not in index | 0.0%#163 | – | Artificial Analysis | |
| Terminal-Bench 2.1 (AA)not in index | 12.0%#192 | – | Artificial Analysis | |
| LiveCodeBench (AA)not in index | 78.8%#32 | 52.4%#131 | Artificial Analysis |
Agents & tools
| Benchmark | thinking | not stated | Source | Trend |
|---|---|---|---|---|
| Terminal-Bench Hard | 13.6%#203 | 15.2%#195 | Artificial Analysis | |
| τ²-Bench Telecom (AA)not in index | 53.2%#192 | 33.3%#246 | Artificial Analysis | |
| BFCL Overalleffort not stated | – | 52.1%#21 | Berkeley Function Calling Leaderboard | |
| τ³-Banking (AA) | 7.8%#161 | – | Artificial Analysis | |
| Vending-Bench 2not in index | -11.3#67 | – | Andon Labs | |
| AutomationBenchnot in index | 2.1%#192 | – | Artificial Analysis | |
| GDP.pdf (AA)not in index | 4.8%#179 | – | Artificial Analysis | |
| BrowseCompnot in index | 0.5%#29 | – | OpenAI | |
| FACTS Searchnot in indexeffort not stated | – | 53.4%#22 |
Maths
| Benchmark | thinking | not stated | Source | Trend |
|---|---|---|---|---|
| OTIS Mock AIME | 86.7%#88 | – | Epoch AI Benchmarking Hub | |
| Omni-MATH (HELM)effort not stated | – | 71.8%#3 | HELM Capabilities | |
| Arena Maths | 1396#156 | 1417#127 | Arena (LMArena) |
Knowledge
| Benchmark | thinking | not stated | Source | Trend |
|---|---|---|---|---|
| SimpleQA Verified | 40.4%#49 | – | Epoch AI Benchmarking Hub | |
| PRBench Financeeffort not stated | – | 39.1%#30 | Scale AI SEAL | |
| PRBench Legaleffort not stated | – | 38.3%#36 | Scale AI SEAL | |
| MultiNRCpossible contamination | 27.1%#28 | – | Scale AI SEAL | |
| MMLU-Pro (HELM)not in indexeffort not stated | – | 84.4%#9 | HELM Capabilities |
Instruction following
| Benchmark | thinking | not stated | Source | Trend |
|---|---|---|---|---|
| IFBench | 51.2%#169 | 46.0%#199 | Artificial Analysis | |
| IFEval (HELM)effort not stated | – | 83.5%#31 | HELM Capabilities | |
| WildBench (HELM)effort not stated | – | 86.6%#1 | HELM Capabilities | |
| Arena Instruction Followingnot in index | 1385#155 | 1414#120 | Arena (LMArena) |
Human preference
| Benchmark | thinking | not stated | Source | Trend |
|---|---|---|---|---|
| Arena Text | 1400#152 | 1422#123 | Arena (LMArena) | |
| Arena Creative Writingnot in index | 1370#150 | 1378#138 | Arena (LMArena) | |
| Arena Multi-turnnot in index | 1393#166 | 1436#107 | Arena (LMArena) | |
| Arena Longer Queriesnot in index | 1402#160 | 1433#119 | Arena (LMArena) | |
| Lech Mazur Creative Writingnot in index | 82.4%#11 | – | Lech Mazur |
Long context
| Benchmark | thinking | not stated | Source | Trend |
|---|---|---|---|---|
| Fiction.LiveBench 120k | 68.8%#8 | – | Fiction.live | |
| AA-LCR v1.1 | 72.0%#199 | 33.9%#391 | Artificial Analysis | |
| MLCRnot in index | 8.9%#126 | – | Artificial Analysis |
Composite
| Benchmark | thinking | not stated | Source | Trend |
|---|---|---|---|---|
| Epoch Capabilities Indexnot in index | 143.8#93 | 138.8#120 | Epoch AI Benchmarking Hub | |
| AA Intelligence Index v4.3.2not in index | 12.7#335 | 12#349 | Artificial Analysis | |
| AA Progress Index v1not in index | 237.5#292 | 215.2#355 | Artificial Analysis | |
| HELM Capabilities meannot in indexeffort not stated | – | 79.8%#6 | HELM Capabilities |
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 |
|---|---|---|---|---|---|---|
| Parasail | 36 tok/s | 0.42 s | $0.140 | $0.800 | 131k | fp8 |
| Venice | 24 tok/s | 0.64 s | $0.150 | $0.750 | 128k | fp8 |
| DeepInfra | 19 tok/s | 0.43 s | $0.090 | $0.550 | 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 $0.200 per 1M. Reasoning tokens are not modelled.
| Workload | Tokens in / out | Cost | With caching | Time |
|---|---|---|---|---|
| Chat reply | 400 / 300 | $0.0003 | $0.0003 | 22.2 s |
| Summarize a 30-page report | 12,000 / 600 | $0.0030 | $0.0029 | 25.0 s |
| Code edit | 6,000 / 1,500 | $0.0025 | $0.0025 | 33.4 s |
| Agentic coding session | 60,000 / 4,000 | $0.016 | $0.015 | 56.8 s |
| Structured extraction | 2,000 / 200 | $0.0006 | $0.0006 | 21.3 s |
More from Alibaba
See also
Questions people ask
- What are Qwen3 235B A22B 2507's benchmark scores?
- As of 29 Oct 2026, Qwen3 235B A22B 2507 (thinking effort) scores 51.9 ±4.0 on the BenchLeader Index, strongest in long context (60) and weakest in agents & tools (39). That rests on 15 independent benchmark results, and every score on this page links to its publisher.
- How does Qwen3 235B A22B 2507 rank?
- Qwen3 235B A22B 2507 (thinking effort) is #286 of 581 ranked configurations on 29 Oct 2026. 266 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 69.2.
- How much does Qwen3 235B A22B 2507 cost per 1M tokens?
- As of 29 Oct 2026, Qwen3 235B A22B 2507 costs $0.210 per 1M input tokens and $0.840 per 1M output tokens, $0.368 blended (three input tokens per output token). The list price last changed on 24 Oct 2026.
- When was Qwen3 235B A22B 2507 released?
- Alibaba released Qwen3 235B A22B 2507 on 22 Jul 2025. BenchLeader has results for it at 2 reasoning-effort settings.
Data as of 29 Oct 2026. Compare these configurations.
Cite as: BenchLeader, “Qwen3 235B A22B 2507: benchmarks, pricing, speed and rank”, data as of 29 Oct 2026, https://www.benchleader.com/models/qwen3-235b-a22b-2507.