GPT-4.1 Mini vs Qwen3 235B A22B Instruct 2507
Qwen3 235B A22B Instruct 2507 leads GPT-4.1 Mini on the Artificial Analysis Intelligence Index, 12.0 to 10.2, and Qwen3 235B A22B Instruct 2507 costs 2.1x less per output token ($0.775 vs $1.60 per million). Updated October 22, 2026; 23 benchmarks have a result for both.
GPT-4.1 Mini or Qwen3 235B A22B Instruct 2507: which to pick
Neither model has a coding score yet, so the two cannot be ranked on coding.
Pick GPT-4.1 Mini for a larger context window (1.05M against 262K tokens) and work measured by LiveCodeBench (Artificial Analysis run) or Terminal-Bench Hard (Artificial Analysis run), where it leads.
- Shared results
- 23
- GPT-4.1 Mini only
- 65
- Qwen3 235B A22B Instruct 2507 only
- 1
- Like-for-like categories
- 8 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
Neither has a coding score yet.
One side not measured
Intelligence
AA Intelligence Index
Qwen3 235B A22B Instruct 2507
Qwen3 235B A22B Instruct 2507 leads, 12.0 against 10.2.
Same board, both measured
Reasoning
GPQA Diamond
No call
Only GPT-4.1 Mini has a GPQA Diamond result, 67.9%.
Directional only
Math
AIME
No call
Only GPT-4.1 Mini has a AIME result, 49.4%.
Directional only
Knowledge
MMLU Pro
No call
Only GPT-4.1 Mini has a MMLU Pro result, 77.2%.
Directional only
Web design
Design Arena: websites
Qwen3 235B A22B Instruct 2507
Qwen3 235B A22B Instruct 2507 leads, 1058 against 997.
Same board, both measured
Long documents
Listed context window
GPT-4.1 Mini
GPT-4.1 Mini takes 1.05M tokens against 262K.
Listed by OpenRouter
Everyday cost
1K input + 500 output tokens
Qwen3 235B A22B Instruct 2507
GPT-4.1 Mini $0.0012, Qwen3 235B A22B Instruct 2507 $0.0005 per turn.
List prices
Cost per solved task
One SWE-bench Verified task, solved
No call
Only GPT-4.1 Mini has the number, $1.818.
Directional only
At a glance
- Coding index
- GPT-4.1 Mini not benchmarked, Qwen3 235B A22B Instruct 2507 not benchmarked.
- Widest category
- Math, 44.7 points apart on average over 2 shared benchmarks.
- Widest single gap
- AA Math Index: GPT-4.1 Mini 46.3 against Qwen3 235B A22B Instruct 2507 91.0, 44.7 points apart on a 0 to 100 scale.
- Output price
- GPT-4.1 Mini $1.60 against Qwen3 235B A22B Instruct 2507 $0.775 per million output tokens, 2.1x apart.
- Context window
- GPT-4.1 Mini takes 1.05M tokens, Qwen3 235B A22B Instruct 2507 262K.
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 | GPT-4.1 Mini | Bars | Qwen3 235B A22B Instruct 2507 | Ties |
|---|---|---|---|---|
| Coding | 0 | 3 | ||
| Intelligence | 0 | 1 | ||
| Agentic work | 0 | 2 | ||
| Reasoning | 0 | 4 | ||
| Math | 0 | 2 | ||
| Knowledge | 0 | 1 | ||
| Web design | 1 | 5 | ||
| General | 0 | 4 | ||
| All shared boards | 1 | 22 |
CodingLike-for-likeSix of these boards feed the Vibe Coding Index03
| Benchmark | GPT-4.1 Mini | Qwen3 235B A22B Instruct 2507 | Leader |
|---|---|---|---|
| Aider polyglot | 32.4% | Not measured | Only GPT-4.1 Mini measured |
| SWE-bench Verified | 23.9% | Not measured | Only GPT-4.1 Mini measured |
| AA Coding Index | 20.2 | Not measured | Only GPT-4.1 Mini measured |
| LiveCodeBench (Vals AI) | 58.2% | Not measured | Only GPT-4.1 Mini measured |
| LiveCodeBench (Artificial Analysis run) | 48.3% | 78.8% | Qwen3 235B A22B Instruct 2507 |
| SciCode | Not measured | 41.4% | Only Qwen3 235B A22B Instruct 2507 measured |
| Terminal-Bench Hard (Artificial Analysis run) | 7.6% | 15.2% | Qwen3 235B A22B Instruct 2507 |
| WeirdML | 37.6% | 38.7% | Qwen3 235B A22B Instruct 2507 |
| LMArena Text: coding | 1367 | Not measured | Only GPT-4.1 Mini measured |
| LMArena Text: software and IT | 1360 | Not measured | Only GPT-4.1 Mini measured |
| Neither model has a result on the other 33 boards here | |||
IntelligenceLike-for-likePublic boards, shown as published, kept out of the index01
| Benchmark | GPT-4.1 Mini | Qwen3 235B A22B Instruct 2507 | Leader |
|---|---|---|---|
| AA Intelligence Index | 10.2 | 12.0 | Qwen3 235B A22B Instruct 2507 |
| Epoch Capabilities Index | 134.9 | Not measured | Only GPT-4.1 Mini measured |
Agentic workLike-for-likePublic boards, shown as published, kept out of the index02
| Benchmark | GPT-4.1 Mini | Qwen3 235B A22B Instruct 2507 | Leader |
|---|---|---|---|
| τ²-Bench Telecom | 52.9% | 53.2% | Qwen3 235B A22B Instruct 2507 |
| Berkeley Function Calling Leaderboard (BFCL) | 50.4% | 52.1% | Qwen3 235B A22B Instruct 2507 |
| Neither model has a result on the other 40 boards here | |||
ReasoningLike-for-likePublic boards, shown as published, kept out of the index04
| Benchmark | GPT-4.1 Mini | Qwen3 235B A22B Instruct 2507 | Leader |
|---|---|---|---|
| GPQA Diamond | 67.9% | Not measured | Only GPT-4.1 Mini measured |
| MMMU Pro (Vals AI) | 70.5% | Not measured | Only GPT-4.1 Mini measured |
| GPQA Diamond (Artificial Analysis run) | 66.4% | 79.0% | Qwen3 235B A22B Instruct 2507 |
| Humanity's Last Exam (Artificial Analysis run) | 5.0% | 15.9% | Qwen3 235B A22B Instruct 2507 |
| GPQA Diamond (Epoch AI run) | 65.8% | Not measured | Only GPT-4.1 Mini measured |
| ARC-AGI | 3.5% | 11.0% | Qwen3 235B A22B Instruct 2507 |
| ARC-AGI-2 | 0.0% | 1.3% | Qwen3 235B A22B Instruct 2507 |
| Chess puzzles | 1.0% | Not measured | Only GPT-4.1 Mini measured |
| Mystery game puzzles | 7.0% | Not measured | Only GPT-4.1 Mini measured |
| LMArena Text: hard prompts | 1349 | Not measured | Only GPT-4.1 Mini measured |
| LMArena Text: expert | 1338 | Not measured | Only GPT-4.1 Mini measured |
| LMArena Text: hard prompts in English | 1362 | Not measured | Only GPT-4.1 Mini measured |
| LMArena Vision: diagrams | 1182 | Not measured | Only GPT-4.1 Mini measured |
| LMArena Vision: homework | 1214 | Not measured | Only GPT-4.1 Mini measured |
| Kagi LLM Benchmark | 48.6% | Not measured | Only GPT-4.1 Mini measured |
| Neither model has a result on the other 8 boards here | |||
MathLike-for-likePublic boards, shown as published, kept out of the index02
| Benchmark | GPT-4.1 Mini | Qwen3 235B A22B Instruct 2507 | Leader |
|---|---|---|---|
| AIME | 49.4% | Not measured | Only GPT-4.1 Mini measured |
| MATH 500 (Vals AI) | 88.0% | Not measured | Only GPT-4.1 Mini measured |
| MGSM (Vals AI) | 87.8% | Not measured | Only GPT-4.1 Mini measured |
| AA Math Index | 46.3 | 91.0 | Qwen3 235B A22B Instruct 2507 |
| AIME 2025 (Artificial Analysis run) | 46.3% | 91.0% | Qwen3 235B A22B Instruct 2507 |
| OTIS Mock AIME | 44.7% | Not measured | Only GPT-4.1 Mini measured |
| FrontierMath, tiers 1 to 3 | 6.7% | Not measured | Only GPT-4.1 Mini measured |
| MATH level 5 | 87.3% | Not measured | Only GPT-4.1 Mini measured |
| LMArena Text: math | 1343 | Not measured | Only GPT-4.1 Mini measured |
| LMArena Text: mathematical work | 1347 | Not measured | Only GPT-4.1 Mini measured |
| Neither model has a result on the other 4 boards here | |||
KnowledgeLike-for-likePublic boards, shown as published, kept out of the index01
| Benchmark | GPT-4.1 Mini | Qwen3 235B A22B Instruct 2507 | Leader |
|---|---|---|---|
| MMLU Pro | 77.2% | Not measured | Only GPT-4.1 Mini measured |
| LegalBench (Vals AI) | 78.0% | Not measured | Only GPT-4.1 Mini measured |
| CorpFin (Vals AI) | 57.9% | Not measured | Only GPT-4.1 Mini measured |
| TaxEval (Vals AI) | 71.9% | Not measured | Only GPT-4.1 Mini measured |
| MortgageTax (Vals AI) | 65.5% | Not measured | Only GPT-4.1 Mini measured |
| MedQA (Vals AI) | 84.6% | Not measured | Only GPT-4.1 Mini measured |
| MMLU Pro (Artificial Analysis run) | 78.1% | 84.3% | Qwen3 235B A22B Instruct 2507 |
| SimpleQA Verified | 12.7% | Not measured | Only GPT-4.1 Mini measured |
| LMArena Text: science | 1328 | Not measured | Only GPT-4.1 Mini measured |
| LMArena Text: medicine and healthcare | 1324 | Not measured | Only GPT-4.1 Mini measured |
| LMArena Text: legal and government | 1343 | Not measured | Only GPT-4.1 Mini measured |
| LMArena Text: business and finance | 1322 | Not measured | Only GPT-4.1 Mini measured |
| PRBench Finance | 30.4% | Not measured | Only GPT-4.1 Mini measured |
| PRBench Legal | 30.4% | Not measured | Only GPT-4.1 Mini measured |
| Neither model has a result on the other 6 boards here | |||
Web designLike-for-likePublic boards, shown as published, kept out of the index15
| Benchmark | GPT-4.1 Mini | Qwen3 235B A22B Instruct 2507 | Leader |
|---|---|---|---|
| Design Arena: websites | 997 | 1058 | Qwen3 235B A22B Instruct 2507 |
| Design Arena: UI components | 959 | 963 | Qwen3 235B A22B Instruct 2507 |
| Design Arena: game dev | 1070 | 951 | GPT-4.1 Mini |
| Design Arena: data viz | 1034 | 1070 | Qwen3 235B A22B Instruct 2507 |
| Design Arena: 3D | 836 | 994 | Qwen3 235B A22B Instruct 2507 |
| Design Arena: all code categories | 993 | 1038 | Qwen3 235B A22B Instruct 2507 |
| Neither model has a result on the other 3 boards here | |||
GeneralLike-for-likePublic boards, shown as published, kept out of the index04
| Benchmark | GPT-4.1 Mini | Qwen3 235B A22B Instruct 2507 | Leader |
|---|---|---|---|
| IFBench | 38.3% | 51.2% | Qwen3 235B A22B Instruct 2507 |
| Long context reasoning (AA-LCR) | 44.0% | 72.0% | Qwen3 235B A22B Instruct 2507 |
| DTBench | 68.8% | 78.4% | Qwen3 235B A22B Instruct 2507 |
| LMCA | 21.1% | 23.5% | Qwen3 235B A22B Instruct 2507 |
| Fiction.LiveBench | 44.4% | Not measured | Only GPT-4.1 Mini measured |
| LMArena Text | 1341 | Not measured | Only GPT-4.1 Mini measured |
| LMArena Text: instruction following | 1333 | Not measured | Only GPT-4.1 Mini measured |
| LMArena Text: creative writing | 1299 | Not measured | Only GPT-4.1 Mini measured |
| LMArena Vision | 1181 | Not measured | Only GPT-4.1 Mini measured |
| LMArena Text: English | 1356 | Not measured | Only GPT-4.1 Mini measured |
| LMArena Text: non-English | 1318 | Not measured | Only GPT-4.1 Mini measured |
| LMArena Text: Chinese | 1329 | Not measured | Only GPT-4.1 Mini measured |
| LMArena Text: German | 1350 | Not measured | Only GPT-4.1 Mini measured |
| LMArena Text: Japanese | 1290 | Not measured | Only GPT-4.1 Mini measured |
| LMArena Text: Korean | 1298 | Not measured | Only GPT-4.1 Mini measured |
| LMArena Text: French | 1358 | Not measured | Only GPT-4.1 Mini measured |
| LMArena Text: Spanish | 1319 | Not measured | Only GPT-4.1 Mini measured |
| LMArena Text: Russian | 1324 | Not measured | Only GPT-4.1 Mini measured |
| LMArena Text: Polish | 1326 | Not measured | Only GPT-4.1 Mini measured |
| LMArena Text: multi-turn | 1354 | Not measured | Only GPT-4.1 Mini measured |
| LMArena Text: longer queries | 1344 | Not measured | Only GPT-4.1 Mini measured |
| LMArena Text: writing and language | 1319 | Not measured | Only GPT-4.1 Mini measured |
| LMArena Text: entertainment and media | 1300 | Not measured | Only GPT-4.1 Mini measured |
| LMArena Vision: English | 1191 | Not measured | Only GPT-4.1 Mini measured |
| LMArena Vision: Chinese | 1186 | Not measured | Only GPT-4.1 Mini measured |
| LMArena Vision: reading text | 1188 | Not measured | Only GPT-4.1 Mini measured |
| LMArena Vision: humor | 1182 | Not measured | Only GPT-4.1 Mini measured |
| LMArena Vision: captioning | 1174 | Not measured | Only GPT-4.1 Mini measured |
| LMArena Vision: recognizing things | 1187 | Not measured | Only GPT-4.1 Mini measured |
| LMArena Vision: creative writing | 1182 | Not measured | Only GPT-4.1 Mini measured |
| Neither model has a result on the other 18 boards here | |||
Image generationNot comparablePublic boards, shown as published, kept out of the index0 boards, none shared
| Benchmark | GPT-4.1 Mini | Qwen3 235B A22B Instruct 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 Instruct 2507 costs less
Repository review
50K input + 3K output tokens
Qwen3 235B A22B Instruct 2507 costs less
Agent loop
200K input + 10K output tokens
Qwen3 235B A22B Instruct 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 | OpenAI GPT-4.1 Mini | Qwen3 235B A22B Instruct 2507 |
|---|---|---|
| Maker | OpenAI | Qwen |
| Weights | No public weights listed | Open weights |
| Released | April 14, 2025 | July 21, 2025 |
| Context window | 1.05M | 262K |
| Input $/Mtok | $0.40 | $0.145 |
| Output $/Mtok | $1.60 | $0.775 |
| Coding index | Not benchmarked | Not benchmarked |
| Independent measurements behind it | 0 | 0 |
| Cost per solved task | $1.818 | Not computable |
| Source | OpenRouter listing | OpenRouter listing |
Questions
Is GPT-4.1 Mini better than Qwen3 235B A22B Instruct 2507 for coding?
Neither model has a coding score yet, so the two cannot be ranked on coding. Pick GPT-4.1 Mini for a larger context window (1.05M against 262K tokens) and work measured by LiveCodeBench (Artificial Analysis run) or Terminal-Bench Hard (Artificial Analysis run), where it leads.
Which is cheaper, GPT-4.1 Mini or Qwen3 235B A22B Instruct 2507?
GPT-4.1 Mini costs $0.40 per million input tokens and $1.60 per million output tokens; Qwen3 235B A22B Instruct 2507 costs $0.145 and $0.775. Cost per solved task is known for GPT-4.1 Mini only.
Which has the bigger context window, GPT-4.1 Mini or Qwen3 235B A22B Instruct 2507?
GPT-4.1 Mini, with 1.05M tokens against 262K.
How many benchmarks is the GPT-4.1 Mini vs Qwen3 235B A22B Instruct 2507 comparison based on?
23 benchmarks have a result for both models as of October 22, 2026. GPT-4.1 Mini's coding index rests on 0 independent benchmarks and Qwen3 235B A22B Instruct 2507's on 0. Where only one model has a result, the page says so instead of naming a winner.