Claude Sonnet 4.6 vs GPT-5.2
Claude Sonnet 4.6 leads GPT-5.2 on coding, 57.3 to 55.9 on the Vibe Coding Index, at about the same output price ($15.00 vs $14.00 per million). Updated October 22, 2026; 90 benchmarks have a result for both.
Claude Sonnet 4.6 or GPT-5.2: which to pick
Claude Sonnet 4.6 and GPT-5.2 are level for coding, 57.3 against 55.9 on the Vibe Coding Index, so price and context decide.
Pick GPT-5.2 for output tokens 1.1x cheaper ($14.00 vs $15.00 per million) and work measured by LiveCodeBench (Vals AI) or Terminal-Bench, where it leads. Pick Claude Sonnet 4.6 for a lower cost per solved task ($1.68 against $1.91) and a larger context window (1M against 400K tokens).
- Shared results
- 90
- Claude Sonnet 4.6 only
- 28
- GPT-5.2 only
- 24
- 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
Level: 57.3 against 55.9, inside 2 points.
5 and 3 benchmarks behind the scores
Intelligence
AA Intelligence Index
GPT-5.2
GPT-5.2 leads, 30.4 against 30.1.
Same board, both measured
Agentic work
AA Agentic Index
No call
Only Claude Sonnet 4.6 has a AA Agentic Index result, 31.8.
Directional only
Reasoning
GPQA Diamond
GPT-5.2
GPT-5.2 leads, 91.7% against 85.6%.
Same board, both measured
Math
AIME
GPT-5.2
GPT-5.2 leads, 96.9% against 92.3%.
Same board, both measured
Knowledge
MMLU Pro
Claude Sonnet 4.6
Claude Sonnet 4.6 leads, 87.3% against 86.2%.
Same board, both measured
Web design
Design Arena: websites
Claude Sonnet 4.6
Claude Sonnet 4.6 leads, 1284 against 1195.
Same board, both measured
Long documents
Listed context window
Claude Sonnet 4.6
Claude Sonnet 4.6 takes 1M tokens against 400K.
Listed by OpenRouter
Everyday cost
1K input + 500 output tokens
GPT-5.2
Claude Sonnet 4.6 $0.011, GPT-5.2 $0.0088 per turn.
List prices
Cost per solved task
One SWE-bench Verified task, solved
Claude Sonnet 4.6
Claude Sonnet 4.6 $1.68, GPT-5.2 $1.91.
Published attempt cost and pass rate
At a glance
- Coding index
- Claude Sonnet 4.6 57.3 from 5 benchmarks, GPT-5.2 55.9 from 3 benchmarks.
- Widest category
- Agentic work, 14.8 points apart on average over 8 shared benchmarks.
- Widest single gap
- ProofBench: Claude Sonnet 4.6 45.0% against GPT-5.2 15.0%, 30.0 points apart on a 0 to 100 scale.
- Output price
- Claude Sonnet 4.6 $15.00 against GPT-5.2 $14.00 per million output tokens, 1.1x apart.
- Context window
- Claude Sonnet 4.6 takes 1M tokens, GPT-5.2 400K.
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 | Claude Sonnet 4.6 | Bars | GPT-5.2 | Ties |
|---|---|---|---|---|
| Coding | 7 | 3 | ||
| Intelligence | 0 | 2 | ||
| Agentic work | 7 | 1 | ||
| Reasoning | 7 | 7 | ||
| Math | 3 | 2 | ||
| Knowledge | 7 | 6 | ||
| Web design | 8 | 0 | ||
| General | 24 | 6 | ||
| All shared boards | 63 | 27 |
CodingLike-for-likeSix of these boards feed the Vibe Coding Index73
| Benchmark | Claude Sonnet 4.6 | GPT-5.2 | Leader |
|---|---|---|---|
| SWE-bench Verified | Not measured | 72.8% | Only GPT-5.2 measured |
| SWE-bench Multilingual | Not measured | 66.7% | Only GPT-5.2 measured |
| LiveBench coding | 66.6% | Not measured | Only Claude Sonnet 4.6 measured |
| LMArena WebDev | 1523 | 1417 | Claude Sonnet 4.6 |
| AA Coding Index | 63.0 | Not measured | Only Claude Sonnet 4.6 measured |
| LiveCodeBench (Vals AI) | 82.1% | 85.4% | GPT-5.2 |
| SWE-bench Verified (Vals AI) | 77.4% | 75.8% | Claude Sonnet 4.6 |
| Vibe Code Bench (Vals AI) | 55.8% | 53.5% | Claude Sonnet 4.6 |
| Code migration (Vals AI) | 39.9% | Not measured | Only Claude Sonnet 4.6 measured |
| ProgramBench (Vals AI) | 0.5% | Not measured | Only Claude Sonnet 4.6 measured |
| Terminal-Bench 2.1 (Vals AI) | 57.3% | Not measured | Only Claude Sonnet 4.6 measured |
| SkillsBench (Vals AI) | 49.0% | Not measured | Only Claude Sonnet 4.6 measured |
| LiveCodeBench (Artificial Analysis run) | Not measured | 89.4% | Only GPT-5.2 measured |
| SciCode | 50.1% | Not measured | Only Claude Sonnet 4.6 measured |
| Terminal-Bench Hard (Artificial Analysis run) | 53.0% | 47.0% | Claude Sonnet 4.6 |
| Terminal-Bench | 53.4% | 64.9% | GPT-5.2 |
| FrontierCode | 24.3% | Not measured | Only Claude Sonnet 4.6 measured |
| GSO-Bench | Not measured | 27.5% | Only GPT-5.2 measured |
| WeirdML | 66.1% | 72.2% | GPT-5.2 |
| ALE-Bench | Not measured | 1293.5 | Only GPT-5.2 measured |
| LMArena Text: coding | 1503 | 1442 | Claude Sonnet 4.6 |
| LMArena Text: software and IT | 1491 | 1434 | Claude Sonnet 4.6 |
| LMArena WebDev: React | 1512 | Not measured | Only Claude Sonnet 4.6 measured |
| LMArena WebDev: HTML | 1527 | 1427 | Claude Sonnet 4.6 |
| LMArena WebDev: from an image | 1529 | Not measured | Only Claude Sonnet 4.6 measured |
| Neither model has a result on the other 14 boards here | |||
IntelligenceLike-for-likePublic boards, shown as published, kept out of the index02
| Benchmark | Claude Sonnet 4.6 | GPT-5.2 | Leader |
|---|---|---|---|
| AA Intelligence Index | 30.1 | 30.4 | GPT-5.2 |
| Epoch Capabilities Index | 152.8 | 153.7 | GPT-5.2 |
Agentic workLike-for-likePublic boards, shown as published, kept out of the index71
| Benchmark | Claude Sonnet 4.6 | GPT-5.2 | Leader |
|---|---|---|---|
| AA Agentic Index | 31.8 | Not measured | Only Claude Sonnet 4.6 measured |
| Design Arena: agents building web apps | 1190 | 1086 | Claude Sonnet 4.6 |
| Design Arena: agents building full-stack apps | 1213 | 1043 | Claude Sonnet 4.6 |
| Design Arena: agents building mobile apps | 1212 | 1093 | Claude Sonnet 4.6 |
| Design Arena: agents building Android apps | 1197 | 1059 | Claude Sonnet 4.6 |
| Design Arena: agents building Godot games | 1226 | 1142 | Claude Sonnet 4.6 |
| Design Arena: agents building games | 1177 | Not measured | Only Claude Sonnet 4.6 measured |
| LiveBench research | 63.6% | Not measured | Only Claude Sonnet 4.6 measured |
| LiveBench browser | 68.0% | Not measured | Only Claude Sonnet 4.6 measured |
| LiveBench long agentic | 23.1% | Not measured | Only Claude Sonnet 4.6 measured |
| Finance Agent (Vals AI) | 51.0% | Not measured | Only Claude Sonnet 4.6 measured |
| Tax Agent Bench (Vals AI) | 25.8% | Not measured | Only Claude Sonnet 4.6 measured |
| Legal Research Bench (Vals AI) | 38.5% | Not measured | Only Claude Sonnet 4.6 measured |
| Excel Modeling Benchmark (Vals AI) | 60.2% | Not measured | Only Claude Sonnet 4.6 measured |
| τ²-Bench Telecom | 79.5% | 84.8% | GPT-5.2 |
| APEX-Agents | 43.0% | Not measured | Only Claude Sonnet 4.6 measured |
| METR time horizons | Not measured | 75.3% | Only GPT-5.2 measured |
| Remote Labor Index | Not measured | 2.5% | Only GPT-5.2 measured |
| GDPval | Not measured | 49.7% | Only GPT-5.2 measured |
| OSWorld 2.0 | 9.3% | Not measured | Only Claude Sonnet 4.6 measured |
| DeepResearch Bench | 54.9% | 41.1% | Claude Sonnet 4.6 |
| OSWorld | 72.1% | Not measured | Only Claude Sonnet 4.6 measured |
| Vending-Bench 2 | $7,204 | $3,591 | Claude Sonnet 4.6 |
| Berkeley Function Calling Leaderboard (BFCL) | Not measured | 55.9% | Only GPT-5.2 measured |
| Neither model has a result on the other 17 boards here | |||
ReasoningLike-for-likePublic boards, shown as published, kept out of the index77
| Benchmark | Claude Sonnet 4.6 | GPT-5.2 | Leader |
|---|---|---|---|
| GPQA Diamond | 85.6% | 91.7% | GPT-5.2 |
| LiveBench reasoning | 86.4% | Not measured | Only Claude Sonnet 4.6 measured |
| MMMU Pro (Vals AI) | 83.6% | 86.7% | GPT-5.2 |
| GPQA Diamond (Artificial Analysis run) | 87.5% | 90.3% | GPT-5.2 |
| Humanity's Last Exam (Artificial Analysis run) | 33.6% | 37.7% | GPT-5.2 |
| GPQA Diamond (Epoch AI run) | 87.4% | 91.4% | GPT-5.2 |
| ARC-AGI | 86.5% | 86.2% | Claude Sonnet 4.6 |
| ARC-AGI-2 | 60.4% | 52.9% | Claude Sonnet 4.6 |
| SimpleBench | Not measured | 45.8% | Only GPT-5.2 measured |
| Chess puzzles | 23.0% | 48.0% | GPT-5.2 |
| Mystery game puzzles | 16.0% | 23.0% | GPT-5.2 |
| EBR-bench | Not measured | 23.0% | Only GPT-5.2 measured |
| VPCT | Not measured | 84.0% | Only GPT-5.2 measured |
| LMArena Text: hard prompts | 1485 | 1428 | Claude Sonnet 4.6 |
| LMArena Text: expert | 1500 | 1445 | Claude Sonnet 4.6 |
| LMArena Text: hard prompts in English | 1493 | 1431 | Claude Sonnet 4.6 |
| LMArena Vision: diagrams | 1306 | 1265 | Claude Sonnet 4.6 |
| LMArena Vision: homework | 1311 | 1276 | Claude Sonnet 4.6 |
| Kagi LLM Benchmark | Not measured | 73.3% | Only GPT-5.2 measured |
| Neither model has a result on the other 3 boards here | |||
MathLike-for-likePublic boards, shown as published, kept out of the index32
| Benchmark | Claude Sonnet 4.6 | GPT-5.2 | Leader |
|---|---|---|---|
| AIME | 92.3% | 96.9% | GPT-5.2 |
| MGSM (Vals AI) | Not measured | 94.0% | Only GPT-5.2 measured |
| AA Math Index | Not measured | 99.0 | Only GPT-5.2 measured |
| AIME 2025 (Artificial Analysis run) | Not measured | 99.0% | Only GPT-5.2 measured |
| OTIS Mock AIME | 85.8% | 96.1% | GPT-5.2 |
| FrontierMath, tiers 1 to 3 | Not measured | 67.4% | Only GPT-5.2 measured |
| FrontierMath, tier 4 | Not measured | 31.7% | Only GPT-5.2 measured |
| ProofBench | 45.0% | 15.0% | Claude Sonnet 4.6 |
| LMArena Text: math | 1462 | 1440 | Claude Sonnet 4.6 |
| LMArena Text: mathematical work | 1483 | 1442 | Claude Sonnet 4.6 |
| Neither model has a result on the other 3 boards here | |||
KnowledgeLike-for-likePublic boards, shown as published, kept out of the index76
| Benchmark | Claude Sonnet 4.6 | GPT-5.2 | Leader |
|---|---|---|---|
| MMLU Pro | 87.3% | 86.2% | Claude Sonnet 4.6 |
| LegalBench (Vals AI) | 82.1% | 82.8% | GPT-5.2 |
| CaseLaw (Vals AI) | 64.0% | 66.0% | GPT-5.2 |
| CorpFin (Vals AI) | 65.3% | 65.9% | GPT-5.2 |
| TaxEval (Vals AI) | 77.1% | 75.8% | Claude Sonnet 4.6 |
| MortgageTax (Vals AI) | 67.7% | 67.1% | Claude Sonnet 4.6 |
| MedQA (Vals AI) | 92.1% | 94.1% | GPT-5.2 |
| MedCode (Vals AI) | Not measured | 49.8% | Only GPT-5.2 measured |
| MMLU Pro (Artificial Analysis run) | Not measured | 87.4% | Only GPT-5.2 measured |
| SimpleQA Verified | 35.5% | 37.1% | GPT-5.2 |
| LMArena Text: science | 1474 | 1425 | Claude Sonnet 4.6 |
| LMArena Text: medicine and healthcare | 1451 | 1425 | Claude Sonnet 4.6 |
| LMArena Text: legal and government | 1470 | 1427 | Claude Sonnet 4.6 |
| LMArena Text: business and finance | 1464 | 1414 | Claude Sonnet 4.6 |
| FACTS Benchmark Suite | 37.0% | 48.5% | GPT-5.2 |
| Neither model has a result on the other 3 boards here | |||
Web designLike-for-likePublic boards, shown as published, kept out of the index80
| Benchmark | Claude Sonnet 4.6 | GPT-5.2 | Leader |
|---|---|---|---|
| Design Arena: websites | 1284 | 1195 | Claude Sonnet 4.6 |
| Design Arena: UI components | 1269 | 1190 | Claude Sonnet 4.6 |
| Design Arena: game dev | 1258 | 1195 | Claude Sonnet 4.6 |
| Design Arena: data viz | 1278 | 1202 | Claude Sonnet 4.6 |
| Design Arena: 3D | 1234 | 1079 | Claude Sonnet 4.6 |
| Design Arena: all code categories | 1275 | 1170 | Claude Sonnet 4.6 |
| Design Arena: SVG | 1196 | 1139 | Claude Sonnet 4.6 |
| Design Arena: ASCII art | 1235 | 1152 | Claude Sonnet 4.6 |
| LiveBench design | 71.8% | Not measured | Only Claude Sonnet 4.6 measured |
GeneralLike-for-likePublic boards, shown as published, kept out of the index246
| Benchmark | Claude Sonnet 4.6 | GPT-5.2 | Leader |
|---|---|---|---|
| LiveBench data analysis | 76.1% | Not measured | Only Claude Sonnet 4.6 measured |
| Public Benefits Bench (Vals AI) | 62.5% | Not measured | Only Claude Sonnet 4.6 measured |
| SAGE (Vals AI) | 46.6% | 49.3% | GPT-5.2 |
| MedScribe (Vals AI) | Not measured | 84.4% | Only GPT-5.2 measured |
| Vals Multimodal Index | 60.6% | Not measured | Only Claude Sonnet 4.6 measured |
| IFBench | 56.6% | 75.4% | GPT-5.2 |
| Long context reasoning (AA-LCR) | 80.0% | 82.7% | GPT-5.2 |
| DTBench | 89.9% | 90.9% | GPT-5.2 |
| LMCA | 46.5% | 43.9% | Claude Sonnet 4.6 |
| CL-bench | Not measured | 18.2% | Only GPT-5.2 measured |
| Furniture assembly | Not measured | 38.3% | Only GPT-5.2 measured |
| ExploitBench | 23.6% | Not measured | Only Claude Sonnet 4.6 measured |
| LMArena Text | 1458 | 1416 | Claude Sonnet 4.6 |
| LMArena Text: instruction following | 1476 | 1409 | Claude Sonnet 4.6 |
| LMArena Text: creative writing | 1435 | 1376 | Claude Sonnet 4.6 |
| LMArena Vision | 1283 | 1246 | Claude Sonnet 4.6 |
| LMArena Text: English | 1472 | 1422 | Claude Sonnet 4.6 |
| LMArena Text: non-English | 1440 | 1405 | Claude Sonnet 4.6 |
| LMArena Text: Chinese | 1491 | 1447 | Claude Sonnet 4.6 |
| LMArena Text: German | 1428 | 1444 | GPT-5.2 |
| LMArena Text: Japanese | 1420 | 1420 | Claude Sonnet 4.6 |
| LMArena Text: Korean | 1410 | 1374 | Claude Sonnet 4.6 |
| LMArena Text: French | 1465 | 1454 | Claude Sonnet 4.6 |
| LMArena Text: Spanish | 1464 | 1413 | Claude Sonnet 4.6 |
| LMArena Text: Russian | 1440 | 1415 | Claude Sonnet 4.6 |
| LMArena Text: Polish | 1443 | 1426 | Claude Sonnet 4.6 |
| LMArena Text: multi-turn | 1464 | 1419 | Claude Sonnet 4.6 |
| LMArena Text: longer queries | 1479 | 1413 | Claude Sonnet 4.6 |
| LMArena Text: writing and language | 1444 | 1392 | Claude Sonnet 4.6 |
| LMArena Text: entertainment and media | 1431 | 1378 | Claude Sonnet 4.6 |
| LMArena Vision: English | 1298 | 1243 | Claude Sonnet 4.6 |
| LMArena Vision: Chinese | 1307 | 1305 | Claude Sonnet 4.6 |
| LMArena Vision: reading text | 1296 | 1257 | Claude Sonnet 4.6 |
| LMArena Vision: humor | 1248 | 1227 | Claude Sonnet 4.6 |
| LMArena Vision: captioning | Not measured | 1267 | Only GPT-5.2 measured |
| LMArena Vision: recognizing things | 1195 | 1232 | GPT-5.2 |
| LMArena Vision: creative writing | 1272 | 1225 | Claude Sonnet 4.6 |
| LMArena Document | 1482 | 1405 | Claude Sonnet 4.6 |
| 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 | Claude Sonnet 4.6 | GPT-5.2 | 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
GPT-5.2 costs less
Repository review
50K input + 3K output tokens
GPT-5.2 costs less
Agent loop
200K input + 10K output tokens
GPT-5.2 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 | Anthropic Claude Sonnet 4.6 | OpenAI GPT-5.2 |
|---|---|---|
| Maker | Anthropic | OpenAI |
| Weights | No public weights listed | No public weights listed |
| Released | February 17, 2026 | December 10, 2025 |
| Context window | 1M | 400K |
| Input $/Mtok | $3.00 | $1.75 |
| Output $/Mtok | $15.00 | $14.00 |
| Coding index | 57.3 | 55.9 |
| Independent measurements behind it | 5 | 3 |
| Cost per solved task | $1.68 | $1.91 |
| Source | OpenRouter listing | OpenRouter listing |
Questions
Is Claude Sonnet 4.6 better than GPT-5.2 for coding?
Claude Sonnet 4.6 and GPT-5.2 are level for coding, 57.3 against 55.9 on the Vibe Coding Index, so price and context decide. Pick GPT-5.2 for output tokens 1.1x cheaper ($14.00 vs $15.00 per million) and work measured by LiveCodeBench (Vals AI) or Terminal-Bench, where it leads. Pick Claude Sonnet 4.6 for a lower cost per solved task ($1.68 against $1.91) and a larger context window (1M against 400K tokens).
Which is cheaper, Claude Sonnet 4.6 or GPT-5.2?
Claude Sonnet 4.6 costs $3.00 per million input tokens and $15.00 per million output tokens; GPT-5.2 costs $1.75 and $14.00. Per solved SWE-bench Verified task, Claude Sonnet 4.6 costs $1.68 and GPT-5.2 $1.91.
Which has the bigger context window, Claude Sonnet 4.6 or GPT-5.2?
Claude Sonnet 4.6, with 1M tokens against 400K.
How many benchmarks is the Claude Sonnet 4.6 vs GPT-5.2 comparison based on?
90 benchmarks have a result for both models as of October 22, 2026. Claude Sonnet 4.6's coding index rests on 5 independent benchmarks and GPT-5.2's on 3. Where only one model has a result, the page says so instead of naming a winner.