Gemini 3.7 Flash vs Qwen3 235B A22B Thinking 2507
Gemini 3.7 Flash leads Qwen3 235B A22B Thinking 2507 on the Artificial Analysis Intelligence Index, 39.6 to 12.7, at about the same output price ($3.75 vs $3.50 per million). Updated October 23, 2026; 41 benchmarks have a result for both.
Gemini 3.7 Flash or Qwen3 235B A22B Thinking 2507: which to pick
Only Gemini 3.7 Flash has a coding score, 58.0 on the Vibe Coding Index; Qwen3 235B A22B Thinking 2507 has none, so the two cannot be ranked on coding.
Pick Qwen3 235B A22B Thinking 2507 for output tokens 1.1x cheaper ($3.50 vs $3.75 per million).
- Shared results
- 41
- Gemini 3.7 Flash only
- 74
- Qwen3 235B A22B Thinking 2507 only
- 5
- Like-for-like categories
- 7 of 9
Only benchmarks that tested both models count toward a pick; a result only one model has is shown but not compared.
Which one for your work
Each card is decided by one number both models carry. Where only one has it, the card says so and makes no call.
Coding
Vibe Coding Index, six public coding benchmarks
No call
Only Gemini 3.7 Flash has a coding score, so no call is made.
One side not measured
Intelligence
AA Intelligence Index
Gemini 3.7 Flash
Gemini 3.7 Flash leads, 39.6 against 12.7.
Same board, both measured
Agentic work
AA Agentic Index
No call
Only Qwen3 235B A22B Thinking 2507 has a AA Agentic Index result, 1.3.
Directional only
Reasoning
GPQA Diamond
No call
Only Gemini 3.7 Flash has a GPQA Diamond result, 93.9%.
Directional only
Knowledge
MMLU Pro
No call
Only Gemini 3.7 Flash has a MMLU Pro result, 90.1%.
Directional only
Web design
Design Arena: websites
Gemini 3.7 Flash
Gemini 3.7 Flash leads, 1303 against 1052.
Same board, both measured
Long documents
Listed context window
Gemini 3.7 Flash
Gemini 3.7 Flash takes 1.05M tokens against 128K.
Listed by OpenRouter
Everyday cost
1K input + 500 output tokens
Qwen3 235B A22B Thinking 2507
Gemini 3.7 Flash $0.0026, Qwen3 235B A22B Thinking 2507 $0.0022 per turn.
List prices
Cost per solved task
One SWE-bench Verified task, solved
No call
Only Gemini 3.7 Flash has the number, $1.78.
Directional only
At a glance
- Coding index
- Gemini 3.7 Flash 58.0 from 6 benchmarks, Qwen3 235B A22B Thinking 2507 not benchmarked.
- Widest category
- Web design, 40.0 points apart on average over 6 shared benchmarks.
- Widest single gap
- AA Coding Index: Gemini 3.7 Flash 71.5 against Qwen3 235B A22B Thinking 2507 22.1, 49.4 points apart on a 0 to 100 scale.
- Output price
- Gemini 3.7 Flash $3.75 against Qwen3 235B A22B Thinking 2507 $3.50 per million output tokens, 1.1x apart.
- Context window
- Gemini 3.7 Flash takes 1.05M tokens, Qwen3 235B A22B Thinking 2507 128K.
Benchmark by benchmark
First the score: boards each model leads, per category, where both have a result. Then every board, a category to a fold; bars are each result on a 0 to 100 scale, so compare the two in a row, not one row with another.
| Category | Gemini 3.7 Flash | Bars | Qwen3 235B A22B Thinking 2507 | Ties |
|---|---|---|---|---|
| Coding | 3 | 0 | ||
| Intelligence | 1 | 0 | ||
| Reasoning | 5 | 0 | ||
| Math | 3 | 0 | ||
| Knowledge | 5 | 0 | ||
| Web design | 6 | 0 | ||
| General | 18 | 0 | ||
| All shared boards | 41 | 0 |
CodingLike-for-likeSix of these boards feed the Vibe Coding Index30
| Benchmark | Gemini 3.7 Flash | Qwen3 235B A22B Thinking 2507 | Leader |
|---|---|---|---|
| LiveBench coding | 80.5% | Not measured | Only Gemini 3.7 Flash measured |
| LMArena WebDev | 1591 | Not measured | Only Gemini 3.7 Flash measured |
| AA Coding Index | 71.5 | 22.1 | Gemini 3.7 Flash |
| LiveCodeBench (Vals AI) | 88.7% | Not measured | Only Gemini 3.7 Flash measured |
| SWE-bench Verified (Vals AI) | 80.8% | Not measured | Only Gemini 3.7 Flash measured |
| Terminal-Bench 4 (Vals AI) | 12.1% | Not measured | Only Gemini 3.7 Flash measured |
| Vibe Code Bench (Vals AI) | 70.4% | Not measured | Only Gemini 3.7 Flash measured |
| Code migration (Vals AI) | 34.8% | Not measured | Only Gemini 3.7 Flash measured |
| ProgramBench (Vals AI) | 0.0% | Not measured | Only Gemini 3.7 Flash measured |
| IOI (Vals AI) | 67.8% | Not measured | Only Gemini 3.7 Flash measured |
| Terminal-Bench 2.1 (Vals AI) | 77.5% | Not measured | Only Gemini 3.7 Flash measured |
| SkillsBench (Vals AI) | 65.9% | Not measured | Only Gemini 3.7 Flash measured |
| CyberBench (Vals AI) | 43.8% | Not measured | Only Gemini 3.7 Flash measured |
| SciCode | 59.8% | Not measured | Only Gemini 3.7 Flash measured |
| FrontierCode | 43.6% | Not measured | Only Gemini 3.7 Flash measured |
| FrontierSWE | 20.3% | Not measured | Only Gemini 3.7 Flash measured |
| WeirdML | Not measured | 41.0% | Only Qwen3 235B A22B Thinking 2507 measured |
| ALE-Bench | 904.3 | Not measured | Only Gemini 3.7 Flash measured |
| LMArena Text: coding | 1501 | 1424 | Gemini 3.7 Flash |
| LMArena Text: software and IT | 1499 | 1431 | Gemini 3.7 Flash |
| LMArena WebDev: React | 1588 | Not measured | Only Gemini 3.7 Flash measured |
| LMArena WebDev: HTML | 1592 | Not measured | Only Gemini 3.7 Flash measured |
| LMArena WebDev: from an image | 1599 | Not measured | Only Gemini 3.7 Flash measured |
| SciCode | 13.8% | Not measured | Only Gemini 3.7 Flash measured |
| Neither model has a result on the other 19 boards here | |||
IntelligenceLike-for-likePublic boards, shown as published, kept out of the index10
| Benchmark | Gemini 3.7 Flash | Qwen3 235B A22B Thinking 2507 | Leader |
|---|---|---|---|
| AA Intelligence Index | 39.6 | 12.7 | Gemini 3.7 Flash |
| Epoch Capabilities Index | 158.0 | Not measured | Only Gemini 3.7 Flash measured |
Agentic workDirectional onlyPublic boards, shown as published, kept out of the index20 boards, none shared
| Benchmark | Gemini 3.7 Flash | Qwen3 235B A22B Thinking 2507 | Leader |
|---|---|---|---|
| AA Agentic Index | Not measured | 1.3 | Only Qwen3 235B A22B Thinking 2507 measured |
| Design Arena: agents building web apps | 1223 | Not measured | Only Gemini 3.7 Flash measured |
| Design Arena: agents building full-stack apps | 1208 | Not measured | Only Gemini 3.7 Flash measured |
| Design Arena: agents building mobile apps | 1228 | Not measured | Only Gemini 3.7 Flash measured |
| Design Arena: agents building Android apps | 1243 | Not measured | Only Gemini 3.7 Flash measured |
| Design Arena: agents building Godot games | 1264 | Not measured | Only Gemini 3.7 Flash measured |
| Design Arena: agents building games | 1235 | Not measured | Only Gemini 3.7 Flash measured |
| Design Arena: spreadsheets | 1103 | Not measured | Only Gemini 3.7 Flash measured |
| LiveBench research | 76.5% | Not measured | Only Gemini 3.7 Flash measured |
| LiveBench browser | 56.0% | Not measured | Only Gemini 3.7 Flash measured |
| LiveBench long agentic | 28.3% | Not measured | Only Gemini 3.7 Flash measured |
| SREBench (Vals AI) | 4.6% | Not measured | Only Gemini 3.7 Flash measured |
| Finance Agent (Vals AI) | 59.0% | Not measured | Only Gemini 3.7 Flash measured |
| Tax Agent Bench (Vals AI) | 22.8% | Not measured | Only Gemini 3.7 Flash measured |
| Legal Research Bench (Vals AI) | 34.6% | Not measured | Only Gemini 3.7 Flash measured |
| Excel Modeling Benchmark (Vals AI) | 71.3% | Not measured | Only Gemini 3.7 Flash measured |
| Terminal-Bench-Science (Vals AI) | 7.1% | Not measured | Only Gemini 3.7 Flash measured |
| APEX-Agents | 67.8% | Not measured | Only Gemini 3.7 Flash measured |
| Remote Labor Index | 5.0% | Not measured | Only Gemini 3.7 Flash measured |
| Vending-Bench 2 | Not measured | -$11 | Only Qwen3 235B A22B Thinking 2507 measured |
| Neither model has a result on the other 22 boards here | |||
ReasoningLike-for-likePublic boards, shown as published, kept out of the index50
| Benchmark | Gemini 3.7 Flash | Qwen3 235B A22B Thinking 2507 | Leader |
|---|---|---|---|
| GPQA Diamond | 93.9% | Not measured | Only Gemini 3.7 Flash measured |
| LiveBench reasoning | 87.8% | Not measured | Only Gemini 3.7 Flash measured |
| MMMU Pro (Vals AI) | 89.0% | Not measured | Only Gemini 3.7 Flash measured |
| GPQA Diamond (Artificial Analysis run) | 94.5% | Not measured | Only Gemini 3.7 Flash measured |
| Humanity's Last Exam (Artificial Analysis run) | 47.9% | Not measured | Only Gemini 3.7 Flash measured |
| GPQA Diamond (Epoch AI run) | 94.8% | 80.1% | Gemini 3.7 Flash |
| ARC-AGI | 95.5% | Not measured | Only Gemini 3.7 Flash measured |
| ARC-AGI-2 | 84.6% | Not measured | Only Gemini 3.7 Flash measured |
| Chess puzzles | 66.0% | Not measured | Only Gemini 3.7 Flash measured |
| Mystery game puzzles | 37.0% | 9.0% | Gemini 3.7 Flash |
| LMArena Text: hard prompts | 1497 | 1416 | Gemini 3.7 Flash |
| LMArena Text: expert | 1514 | 1463 | Gemini 3.7 Flash |
| LMArena Text: hard prompts in English | 1489 | 1432 | Gemini 3.7 Flash |
| LMArena Vision: diagrams | 1329 | Not measured | Only Gemini 3.7 Flash measured |
| LMArena Vision: homework | 1320 | Not measured | Only Gemini 3.7 Flash measured |
| Neither model has a result on the other 8 boards here | |||
MathLike-for-likePublic boards, shown as published, kept out of the index30
| Benchmark | Gemini 3.7 Flash | Qwen3 235B A22B Thinking 2507 | Leader |
|---|---|---|---|
| OTIS Mock AIME | 97.2% | 86.7% | Gemini 3.7 Flash |
| FrontierMath, tiers 1 to 3 | 71.6% | Not measured | Only Gemini 3.7 Flash measured |
| FrontierMath, tier 4 | 36.6% | Not measured | Only Gemini 3.7 Flash measured |
| ProofBench | 58.0% | Not measured | Only Gemini 3.7 Flash measured |
| LMArena Text: math | 1510 | 1412 | Gemini 3.7 Flash |
| LMArena Text: mathematical work | 1502 | 1435 | Gemini 3.7 Flash |
| MathArena | 25.9% | Not measured | Only Gemini 3.7 Flash measured |
| Neither model has a result on the other 7 boards here | |||
KnowledgeLike-for-likePublic boards, shown as published, kept out of the index50
| Benchmark | Gemini 3.7 Flash | Qwen3 235B A22B Thinking 2507 | Leader |
|---|---|---|---|
| MMLU Pro | 90.1% | Not measured | Only Gemini 3.7 Flash measured |
| LegalBench (Vals AI) | 87.3% | Not measured | Only Gemini 3.7 Flash measured |
| TaxEval (Vals AI) | 74.7% | Not measured | Only Gemini 3.7 Flash measured |
| MortgageTax (Vals AI) | 66.7% | Not measured | Only Gemini 3.7 Flash measured |
| MedCode (Vals AI) | 53.4% | Not measured | Only Gemini 3.7 Flash measured |
| SimpleQA Verified | 69.2% | 40.4% | Gemini 3.7 Flash |
| LMArena Text: science | 1512 | 1433 | Gemini 3.7 Flash |
| LMArena Text: medicine and healthcare | 1495 | 1456 | Gemini 3.7 Flash |
| LMArena Text: legal and government | 1497 | 1415 | Gemini 3.7 Flash |
| LMArena Text: business and finance | 1480 | 1412 | Gemini 3.7 Flash |
| FACTS Benchmark Suite | 71.2% | Not measured | Only Gemini 3.7 Flash measured |
| Neither model has a result on the other 9 boards here | |||
Web designLike-for-likePublic boards, shown as published, kept out of the index60
| Benchmark | Gemini 3.7 Flash | Qwen3 235B A22B Thinking 2507 | Leader |
|---|---|---|---|
| Design Arena: websites | 1303 | 1052 | Gemini 3.7 Flash |
| Design Arena: UI components | 1295 | 942 | Gemini 3.7 Flash |
| Design Arena: game dev | 1318 | 956 | Gemini 3.7 Flash |
| Design Arena: data viz | 1317 | 950 | Gemini 3.7 Flash |
| Design Arena: 3D | 1308 | 998 | Gemini 3.7 Flash |
| Design Arena: all code categories | 1307 | 1032 | Gemini 3.7 Flash |
| Design Arena: SVG | 1284 | Not measured | Only Gemini 3.7 Flash measured |
| Design Arena: ASCII art | 1277 | Not measured | Only Gemini 3.7 Flash measured |
| LiveBench design | 73.2% | Not measured | Only Gemini 3.7 Flash measured |
GeneralLike-for-likePublic boards, shown as published, kept out of the index180
| Benchmark | Gemini 3.7 Flash | Qwen3 235B A22B Thinking 2507 | Leader |
|---|---|---|---|
| LiveBench data analysis | 68.0% | Not measured | Only Gemini 3.7 Flash measured |
| SAGE (Vals AI) | 49.2% | Not measured | Only Gemini 3.7 Flash measured |
| MedScribe (Vals AI) | 83.9% | Not measured | Only Gemini 3.7 Flash measured |
| Long context reasoning (AA-LCR) | 83.0% | Not measured | Only Gemini 3.7 Flash measured |
| DTBench | 96.8% | 80.3% | Gemini 3.7 Flash |
| LMCA | 50.4% | 29.3% | Gemini 3.7 Flash |
| Fiction.LiveBench | Not measured | 75.0% | Only Qwen3 235B A22B Thinking 2507 measured |
| Creative writing (Lech Mazur) | Not measured | 82.4% | Only Qwen3 235B A22B Thinking 2507 measured |
| Furniture assembly | 26.7% | Not measured | Only Gemini 3.7 Flash measured |
| LMArena Text | 1490 | 1415 | Gemini 3.7 Flash |
| LMArena Text: instruction following | 1486 | 1386 | Gemini 3.7 Flash |
| LMArena Text: creative writing | 1489 | 1384 | Gemini 3.7 Flash |
| LMArena Vision | 1314 | Not measured | Only Gemini 3.7 Flash measured |
| LMArena Text: English | 1485 | 1429 | Gemini 3.7 Flash |
| LMArena Text: non-English | 1487 | 1397 | Gemini 3.7 Flash |
| LMArena Text: Chinese | 1553 | 1481 | Gemini 3.7 Flash |
| LMArena Text: German | 1509 | 1408 | Gemini 3.7 Flash |
| LMArena Text: Japanese | 1513 | 1386 | Gemini 3.7 Flash |
| LMArena Text: Korean | 1485 | 1375 | Gemini 3.7 Flash |
| LMArena Text: French | 1500 | Not measured | Only Gemini 3.7 Flash measured |
| LMArena Text: Spanish | 1493 | 1389 | Gemini 3.7 Flash |
| LMArena Text: Russian | 1514 | 1400 | Gemini 3.7 Flash |
| LMArena Text: Polish | 1503 | 1390 | Gemini 3.7 Flash |
| LMArena Text: multi-turn | 1487 | 1406 | Gemini 3.7 Flash |
| LMArena Text: longer queries | 1494 | 1399 | Gemini 3.7 Flash |
| LMArena Text: writing and language | 1491 | 1388 | Gemini 3.7 Flash |
| LMArena Text: entertainment and media | 1478 | 1370 | Gemini 3.7 Flash |
| LMArena Vision: English | 1309 | Not measured | Only Gemini 3.7 Flash measured |
| LMArena Vision: Chinese | 1394 | Not measured | Only Gemini 3.7 Flash measured |
| LMArena Vision: reading text | 1322 | Not measured | Only Gemini 3.7 Flash measured |
| LMArena Vision: humor | 1299 | Not measured | Only Gemini 3.7 Flash measured |
| LMArena Vision: creative writing | 1312 | Not measured | Only Gemini 3.7 Flash measured |
| Neither model has a result on the other 16 boards here | |||
Image generationNot comparablePublic boards, shown as published, kept out of the index0 boards, none shared
| Benchmark | Gemini 3.7 Flash | Qwen3 235B A22B Thinking 2507 | Leader |
|---|---|---|---|
| Neither model has a result on 5 boards here | |||
What each workload costs
Three fixed token mixes at each model's listed price, so per-token rates become a bill.
Chat turn
1K input + 500 output tokens
Qwen3 235B A22B Thinking 2507 costs less
Repository review
50K input + 3K output tokens
Qwen3 235B A22B Thinking 2507 costs less
Agent loop
200K input + 10K output tokens
Qwen3 235B A22B Thinking 2507 costs less
Listed OpenRouter rates, without cache or batch discounts. A workload that does not fit a model's context window would need splitting.
Specification differences
The leading value in a row is marked; a row with a missing value names no leader. Our two rows are marked in red.
| Specification | Google Gemini 3.7 Flash | Qwen3 235B A22B Thinking 2507 |
|---|---|---|
| Maker | Qwen | |
| Weights | No public weights listed | Open weights |
| Released | August 13, 2026 | July 25, 2025 |
| Context window | 1.05M | 128K |
| Input $/Mtok | $0.75 | $0.45 |
| Output $/Mtok | $3.75 | $3.50 |
| Coding index | 58.0 | Not benchmarked |
| Independent measurements behind it | 6 | 0 |
| Cost per solved task | $1.78 | Not computable |
| Source | OpenRouter listing | OpenRouter listing |
Questions
Is Gemini 3.7 Flash better than Qwen3 235B A22B Thinking 2507 for coding?
Only Gemini 3.7 Flash has a coding score, 58.0 on the Vibe Coding Index; Qwen3 235B A22B Thinking 2507 has none, so the two cannot be ranked on coding. Pick Qwen3 235B A22B Thinking 2507 for output tokens 1.1x cheaper ($3.50 vs $3.75 per million).
Which is cheaper, Gemini 3.7 Flash or Qwen3 235B A22B Thinking 2507?
Gemini 3.7 Flash costs $0.75 per million input tokens and $3.75 per million output tokens; Qwen3 235B A22B Thinking 2507 costs $0.45 and $3.50. Cost per solved task is known for Gemini 3.7 Flash only.
Which has the bigger context window, Gemini 3.7 Flash or Qwen3 235B A22B Thinking 2507?
Gemini 3.7 Flash, with 1.05M tokens against 128K.
How many benchmarks is the Gemini 3.7 Flash vs Qwen3 235B A22B Thinking 2507 comparison based on?
41 benchmarks have a result for both models as of October 23, 2026. Gemini 3.7 Flash's coding index rests on 6 independent benchmarks and Qwen3 235B A22B Thinking 2507's on 0. Where only one model has a result, the page says so instead of naming a winner.