GPT-5.5 vs Mistral Large 4
GPT-5.5 leads Mistral Large 4 on coding, 71.3 to 56.4 on the Vibe Coding Index, and Mistral Large 4 costs 7.2x less per output token ($4.18 vs $30.00 per million). Updated October 21, 2026; 74 benchmarks have a result for both.
GPT-5.5 or Mistral Large 4: which to pick
Pick GPT-5.5 for coding: it scores 71.3 on the Vibe Coding Index against 56.4 for Mistral Large 4.
Pick Mistral Large 4 for output tokens 7.2x cheaper ($4.18 vs $30.00 per million) and work measured by LMArena WebDev or Vibe Code Bench (Vals AI), where it leads.
- Shared results
- 74
- GPT-5.5 only
- 60
- Mistral Large 4 only
- 7
- 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
GPT-5.5
GPT-5.5 scores 71.3 against 56.4.
5 benchmarks behind each score
Intelligence
AA Intelligence Index
No call
Tied at 38.4.
Same board, both measured
Agentic work
AA Agentic Index
No call
Only GPT-5.5 has a AA Agentic Index result, 36.4.
Directional only
Reasoning
GPQA Diamond
No call
Only GPT-5.5 has a GPQA Diamond result, 93.2%.
Directional only
Knowledge
MMLU Pro
No call
Only GPT-5.5 has a MMLU Pro result, 88.1%.
Directional only
Web design
Design Arena: websites
Mistral Large 4
Mistral Large 4 leads, 1258 against 1256.
Same board, both measured
Long documents
Listed context window
GPT-5.5
GPT-5.5 takes 1.05M tokens against 1.05M.
Listed by OpenRouter
Everyday cost
1K input + 500 output tokens
Mistral Large 4
GPT-5.5 $0.020, Mistral Large 4 $0.0034 per turn.
List prices
Cost per solved task
One SWE-bench Verified task, solved
No call
Only GPT-5.5 has the number, $1.65.
Directional only
At a glance
- Coding index
- GPT-5.5 71.3 from 5 benchmarks, Mistral Large 4 56.4 from 5 benchmarks.
- Widest category
- Math, 22.5 points apart on average over 6 shared benchmarks.
- Widest single gap
- Surface Evolver Bench: GPT-5.5 88.1% against Mistral Large 4 28.1%, 60.0 points apart on a 0 to 100 scale.
- Output price
- GPT-5.5 $30.00 against Mistral Large 4 $4.18 per million output tokens, 7.2x apart.
- Context window
- GPT-5.5 takes 1.05M tokens, Mistral Large 4 1.05M.
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-5.5 | Bars | Mistral Large 4 | Ties |
|---|---|---|---|---|
| Coding | 8 | 3 | ||
| Intelligence | 1 | 0 | 1 tie | |
| Agentic work | 5 | 2 | ||
| Reasoning | 9 | 0 | ||
| Math | 5 | 1 | ||
| Knowledge | 6 | 0 | ||
| Web design | 5 | 4 | ||
| General | 23 | 1 | ||
| All shared boards | 62 | 11 | 1 tie |
CodingLike-for-likeSix of these boards feed the Vibe Coding Index83
| Benchmark | GPT-5.5 | Mistral Large 4 | Leader |
|---|---|---|---|
| LiveBench coding | 85.1% | 60.0% | GPT-5.5 |
| LMArena WebDev | 1513 | 1543 | Mistral Large 4 |
| AA Coding Index | 74.9 | Not measured | Only GPT-5.5 measured |
| LiveCodeBench (Vals AI) | 85.3% | Not measured | Only GPT-5.5 measured |
| SWE-bench Verified (Vals AI) | 82.6% | Not measured | Only GPT-5.5 measured |
| Terminal-Bench 4 (Vals AI) | Not measured | 27.8% | Only Mistral Large 4 measured |
| Vibe Code Bench (Vals AI) | 69.8% | 78.4% | Mistral Large 4 |
| Code migration (Vals AI) | 45.2% | 30.6% | GPT-5.5 |
| ProgramBench (Vals AI) | 0.5% | Not measured | Only GPT-5.5 measured |
| IOI (Vals AI) | Not measured | 45.3% | Only Mistral Large 4 measured |
| Terminal-Bench 2.1 (Vals AI) | 76.4% | Not measured | Only GPT-5.5 measured |
| SWE-bench Multimodal (Vals AI) | Not measured | 33.1% | Only Mistral Large 4 measured |
| Vibe Code Bench 1-100 (Vals AI) | Not measured | 14.3% | Only Mistral Large 4 measured |
| SkillsBench (Vals AI) | 62.5% | Not measured | Only GPT-5.5 measured |
| SciCode | 56.1% | 54.2% | GPT-5.5 |
| Terminal-Bench Hard (Artificial Analysis run) | 60.6% | Not measured | Only GPT-5.5 measured |
| Terminal-Bench | 84.7% | Not measured | Only GPT-5.5 measured |
| FrontierCode | 43.0% | 39.4% | GPT-5.5 |
| GSO-Bench | 40.2% | Not measured | Only GPT-5.5 measured |
| WeirdML | 84.9% | Not measured | Only GPT-5.5 measured |
| MirrorCode | 10.0% | Not measured | Only GPT-5.5 measured |
| ALE-Bench | 1943.0 | 1213.2 | GPT-5.5 |
| LMArena Text: coding | 1494 | 1478 | GPT-5.5 |
| LMArena Text: software and IT | 1493 | 1467 | GPT-5.5 |
| LMArena WebDev: React | 1506 | 1559 | Mistral Large 4 |
| LMArena WebDev: HTML | 1543 | 1487 | GPT-5.5 |
| LMArena WebDev: from an image | 1532 | Not measured | Only GPT-5.5 measured |
| Neither model has a result on the other 12 boards here | |||
IntelligenceLike-for-likePublic boards, shown as published, kept out of the index10
| Benchmark | GPT-5.5 | Mistral Large 4 | Leader |
|---|---|---|---|
| AA Intelligence Index | 38.4 | 38.4 | Tie |
| Epoch Capabilities Index | 159.7 | 150.0 | GPT-5.5 |
Agentic workLike-for-likePublic boards, shown as published, kept out of the index52
| Benchmark | GPT-5.5 | Mistral Large 4 | Leader |
|---|---|---|---|
| AA Agentic Index | 36.4 | Not measured | Only GPT-5.5 measured |
| Design Arena: agents building web apps | 1117 | Not measured | Only GPT-5.5 measured |
| Design Arena: agents building full-stack apps | 1089 | Not measured | Only GPT-5.5 measured |
| Design Arena: agents building mobile apps | 1143 | Not measured | Only GPT-5.5 measured |
| Design Arena: agents building Android apps | 1163 | Not measured | Only GPT-5.5 measured |
| Design Arena: agents building Godot games | 1173 | Not measured | Only GPT-5.5 measured |
| Design Arena: agents building games | 1179 | Not measured | Only GPT-5.5 measured |
| Design Arena: HTML slides | 1067 | Not measured | Only GPT-5.5 measured |
| Design Arena: PowerPoint slides | 1109 | Not measured | Only GPT-5.5 measured |
| LiveBench research | 61.6% | 28.1% | GPT-5.5 |
| LiveBench browser | 48.0% | 8.0% | GPT-5.5 |
| LiveBench long agentic | 44.6% | 11.8% | GPT-5.5 |
| SREBench (Vals AI) | 3.8% | Not measured | Only GPT-5.5 measured |
| Finance Agent (Vals AI) | 51.8% | 54.7% | Mistral Large 4 |
| Tax Agent Bench (Vals AI) | 21.8% | 30.4% | Mistral Large 4 |
| Legal Research Bench (Vals AI) | 40.4% | 31.7% | GPT-5.5 |
| Harvey's Legal Agent Benchmark (Vals AI) | Not measured | 17.5% | Only Mistral Large 4 measured |
| Excel Modeling Benchmark (Vals AI) | 64.5% | 56.2% | GPT-5.5 |
| BioMysteryBench (Vals AI) | Not measured | 67.0% | Only Mistral Large 4 measured |
| Time Horizon Index (Vals AI) | 8.3% | Not measured | Only GPT-5.5 measured |
| τ²-Bench Telecom | 93.9% | Not measured | Only GPT-5.5 measured |
| APEX-Agents | 55.1% | Not measured | Only GPT-5.5 measured |
| Remote Labor Index | 6.3% | Not measured | Only GPT-5.5 measured |
| OSWorld 2.0 | 13.0% | Not measured | Only GPT-5.5 measured |
| DeepResearch Bench | 54.0% | Not measured | Only GPT-5.5 measured |
| PostTrainBench | 27.2% | Not measured | Only GPT-5.5 measured |
| Vending-Bench 2 | $7,524 | Not measured | Only GPT-5.5 measured |
| Neither model has a result on the other 14 boards here | |||
ReasoningLike-for-likePublic boards, shown as published, kept out of the index90
| Benchmark | GPT-5.5 | Mistral Large 4 | Leader |
|---|---|---|---|
| GPQA Diamond | 93.2% | Not measured | Only GPT-5.5 measured |
| LiveBench reasoning | 89.7% | 83.9% | GPT-5.5 |
| MysteryMechanism (Vals AI) | Not measured | 12.2% | Only Mistral Large 4 measured |
| MMMU Pro (Vals AI) | 88.3% | Not measured | Only GPT-5.5 measured |
| GPQA Diamond (Artificial Analysis run) | 93.5% | Not measured | Only GPT-5.5 measured |
| Humanity's Last Exam (Artificial Analysis run) | 45.8% | 35.0% | GPT-5.5 |
| GPQA Diamond (Epoch AI run) | 90.7% | 88.1% | GPT-5.5 |
| ARC-AGI | 95.0% | Not measured | Only GPT-5.5 measured |
| ARC-AGI-2 | 85.0% | Not measured | Only GPT-5.5 measured |
| SimpleBench | 69.0% | Not measured | Only GPT-5.5 measured |
| Chess puzzles | 63.0% | 33.0% | GPT-5.5 |
| Mystery game puzzles | 56.0% | 17.0% | GPT-5.5 |
| EBR-bench | 34.3% | Not measured | Only GPT-5.5 measured |
| LMArena Text: hard prompts | 1489 | 1444 | GPT-5.5 |
| LMArena Text: expert | 1507 | 1444 | GPT-5.5 |
| LMArena Text: hard prompts in English | 1483 | 1442 | GPT-5.5 |
| LMArena Vision: diagrams | 1325 | 1288 | GPT-5.5 |
| LMArena Vision: homework | 1332 | Not measured | Only GPT-5.5 measured |
| Kagi LLM Benchmark | 88.8% | Not measured | Only GPT-5.5 measured |
| Neither model has a result on the other 3 boards here | |||
MathLike-for-likePublic boards, shown as published, kept out of the index51
| Benchmark | GPT-5.5 | Mistral Large 4 | Leader |
|---|---|---|---|
| OTIS Mock AIME | 84.4% | 98.3% | Mistral Large 4 |
| FrontierMath, tiers 1 to 3 | 85.3% | 58.6% | GPT-5.5 |
| FrontierMath, tier 4 | 72.5% | 26.8% | GPT-5.5 |
| ProofBench | 50.0% | 10.0% | GPT-5.5 |
| LMArena Text: math | 1487 | 1469 | GPT-5.5 |
| LMArena Text: mathematical work | 1496 | 1462 | GPT-5.5 |
| Neither model has a result on the other 7 boards here | |||
KnowledgeLike-for-likePublic boards, shown as published, kept out of the index60
| Benchmark | GPT-5.5 | Mistral Large 4 | Leader |
|---|---|---|---|
| MMLU Pro | 88.1% | Not measured | Only GPT-5.5 measured |
| LegalBench (Vals AI) | 86.5% | Not measured | Only GPT-5.5 measured |
| CaseLaw (Vals AI) | 66.2% | Not measured | Only GPT-5.5 measured |
| CorpFin (Vals AI) | 68.4% | Not measured | Only GPT-5.5 measured |
| TaxEval (Vals AI) | 75.0% | Not measured | Only GPT-5.5 measured |
| MortgageTax (Vals AI) | 68.8% | Not measured | Only GPT-5.5 measured |
| MedCode (Vals AI) | 49.1% | 40.7% | GPT-5.5 |
| SimpleQA Verified | 63.0% | 20.0% | GPT-5.5 |
| LMArena Text: science | 1485 | 1448 | GPT-5.5 |
| LMArena Text: medicine and healthcare | 1465 | 1433 | GPT-5.5 |
| LMArena Text: legal and government | 1491 | 1425 | GPT-5.5 |
| LMArena Text: business and finance | 1478 | 1439 | GPT-5.5 |
| FACTS Benchmark Suite | 66.7% | Not measured | Only GPT-5.5 measured |
| Neither model has a result on the other 5 boards here | |||
Web designLike-for-likePublic boards, shown as published, kept out of the index54
| Benchmark | GPT-5.5 | Mistral Large 4 | Leader |
|---|---|---|---|
| Design Arena: websites | 1256 | 1258 | Mistral Large 4 |
| Design Arena: UI components | 1260 | 1275 | Mistral Large 4 |
| Design Arena: game dev | 1294 | 1226 | GPT-5.5 |
| Design Arena: data viz | 1265 | 1274 | Mistral Large 4 |
| Design Arena: 3D | 1198 | 1253 | Mistral Large 4 |
| Design Arena: all code categories | 1260 | 1259 | GPT-5.5 |
| Design Arena: SVG | 1234 | 1233 | GPT-5.5 |
| Design Arena: ASCII art | 1255 | 1202 | GPT-5.5 |
| LiveBench design | 70.3% | 65.6% | GPT-5.5 |
GeneralLike-for-likePublic boards, shown as published, kept out of the index231
| Benchmark | GPT-5.5 | Mistral Large 4 | Leader |
|---|---|---|---|
| LiveBench data analysis | 81.6% | 76.5% | GPT-5.5 |
| Public Benefits Bench (Vals AI) | 60.9% | Not measured | Only GPT-5.5 measured |
| SAGE (Vals AI) | 51.5% | Not measured | Only GPT-5.5 measured |
| MedScribe (Vals AI) | 86.9% | 80.4% | GPT-5.5 |
| Vals Multimodal Index | 68.1% | Not measured | Only GPT-5.5 measured |
| IFBench | 75.9% | Not measured | Only GPT-5.5 measured |
| Long context reasoning (AA-LCR) | 84.3% | 81.3% | GPT-5.5 |
| DTBench | 96.0% | 88.2% | GPT-5.5 |
| LMCA | 54.3% | 35.8% | GPT-5.5 |
| Surface Evolver Bench | 88.1% | 28.1% | GPT-5.5 |
| Furniture assembly | 44.2% | 20.8% | GPT-5.5 |
| ExploitBench | 47.4% | Not measured | Only GPT-5.5 measured |
| CL-bench Life | 22.2% | Not measured | Only GPT-5.5 measured |
| LMArena Text | 1471 | 1431 | GPT-5.5 |
| LMArena Text: instruction following | 1478 | 1422 | GPT-5.5 |
| LMArena Text: creative writing | 1455 | 1363 | GPT-5.5 |
| LMArena Vision | 1297 | 1272 | GPT-5.5 |
| LMArena Text: English | 1470 | 1433 | GPT-5.5 |
| LMArena Text: non-English | 1467 | 1422 | GPT-5.5 |
| LMArena Text: Chinese | 1534 | 1492 | GPT-5.5 |
| LMArena Text: German | 1480 | Not measured | Only GPT-5.5 measured |
| LMArena Text: Japanese | 1499 | 1415 | GPT-5.5 |
| LMArena Text: Korean | 1460 | Not measured | Only GPT-5.5 measured |
| LMArena Text: French | 1485 | 1492 | Mistral Large 4 |
| LMArena Text: Spanish | 1468 | 1455 | GPT-5.5 |
| LMArena Text: Russian | 1473 | 1417 | GPT-5.5 |
| LMArena Text: Polish | 1465 | Not measured | Only GPT-5.5 measured |
| LMArena Text: multi-turn | 1476 | 1435 | GPT-5.5 |
| LMArena Text: longer queries | 1484 | 1428 | GPT-5.5 |
| LMArena Text: writing and language | 1466 | 1386 | GPT-5.5 |
| LMArena Text: entertainment and media | 1449 | 1378 | GPT-5.5 |
| LMArena Vision: English | 1292 | 1283 | GPT-5.5 |
| LMArena Vision: Chinese | 1360 | Not measured | Only GPT-5.5 measured |
| LMArena Vision: reading text | 1306 | 1284 | GPT-5.5 |
| LMArena Vision: humor | 1293 | Not measured | Only GPT-5.5 measured |
| LMArena Vision: recognizing things | 1231 | Not measured | Only GPT-5.5 measured |
| LMArena Vision: creative writing | 1301 | Not measured | Only GPT-5.5 measured |
| LMArena Document | 1486 | Not measured | Only GPT-5.5 measured |
| Neither model has a result on the other 8 boards here | |||
Image generationNot comparablePublic boards, shown as published, kept out of the index0 boards, none shared
| Benchmark | GPT-5.5 | Mistral Large 4 | 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
Mistral Large 4 costs less
Repository review
50K input + 3K output tokens
Mistral Large 4 costs less
Agent loop
200K input + 10K output tokens
Mistral Large 4 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-5.5 | Mistral Large 4 |
|---|---|---|
| Maker | OpenAI | Mistral AI |
| Weights | No public weights listed | No public weights listed |
| Released | April 24, 2026 | October 6, 2026 |
| Context window | 1.05M | 1.05M |
| Input $/Mtok | $5.00 | $1.36 |
| Output $/Mtok | $30.00 | $4.18 |
| Coding index | 71.3 | 56.4 |
| Independent measurements behind it | 5 | 5 |
| Cost per solved task | $1.65 | Not computable |
| Source | OpenRouter listing | OpenRouter listing |
Questions
Is GPT-5.5 better than Mistral Large 4 for coding?
Pick GPT-5.5 for coding: it scores 71.3 on the Vibe Coding Index against 56.4 for Mistral Large 4. Pick Mistral Large 4 for output tokens 7.2x cheaper ($4.18 vs $30.00 per million) and work measured by LMArena WebDev or Vibe Code Bench (Vals AI), where it leads.
Which is cheaper, GPT-5.5 or Mistral Large 4?
GPT-5.5 costs $5.00 per million input tokens and $30.00 per million output tokens; Mistral Large 4 costs $1.36 and $4.18. Cost per solved task is known for GPT-5.5 only.
Which has the bigger context window, GPT-5.5 or Mistral Large 4?
GPT-5.5, with 1.05M tokens against 1.05M.
How many benchmarks is the GPT-5.5 vs Mistral Large 4 comparison based on?
74 benchmarks have a result for both models as of October 21, 2026. GPT-5.5's coding index rests on 5 independent benchmarks and Mistral Large 4's on 5. Where only one model has a result, the page says so instead of naming a winner.