Model comparison
Claude Sonnet 4.6 vs GPT-5.6 Sol
GPT-5.6 Sol is the stronger model overall, scoring 64.4 to 50.1 on the Noometry Index.
Last verified . 48 shared benchmarks.
Summary
- They share 48 benchmarks with published results for both. Claude Sonnet 4.6 scores higher in 0 categories and GPT-5.6 Sol in 10 categories; 7 gaps are clear of the uncertainty.
- The widest gap is in math, where GPT-5.6 Sol leads 85.5 to 51.9.
- The biggest single-benchmark swing is Chess Puzzles: 13% for Claude Sonnet 4.6 and 64% for GPT-5.6 Sol.
- Claude Sonnet 4.6 is cheaper at $3 / $15 per million input/output tokens, against $4 / $20 for GPT-5.6 Sol.
- GPT-5.6 Sol accepts more context: 1.05M tokens versus 1M.
Side by side
| Claude Sonnet 4.6 | GPT-5.6 Sol | |
|---|---|---|
| Provider | Anthropic | OpenAI |
| Noometry Index | 50.1 | 64.4 |
| Released | 2026-02-17 | 2026-07-09 |
| Weights | Proprietary | Proprietary |
| Context window | 1M | 1.05M |
| Max output | 128K | 128K |
| Input $ / M tokens | $3 | $4 |
| Output $ / M tokens | $15 | $20 |
| Results tracked | 57 | 64 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding GPT-5.6 Sol leads
Claude Sonnet 4.6: 46.0 (#62), GPT-5.6 Sol: 61.1 (#7)
| Benchmark | Claude Sonnet 4.6 | GPT-5.6 Sol |
|---|---|---|
| DeepSWE | 29.9% | 72.7% |
| FrontierCode | 24.3% | 47.5% |
| LMArena WebDev | 1521 | 1619 |
| SciCode | 46.8% | 57.1% |
| WeirdML | 66.1% | 89.4% |
| LMArena Coding | 1504 | 1493 |
| ALE-Bench | 1,327 | 2,177 |
| SWE-bench Verified | 75.2% | — |
| CursorBench | — | 41.7% |
| FrontierSWE | — | 32.2% |
| MirrorCode | — | 20% |
Agentic & Tool Use GPT-5.6 Sol leads
Claude Sonnet 4.6: 39.1 (#26), GPT-5.6 Sol: 50.2 (#7)
| Benchmark | Claude Sonnet 4.6 | GPT-5.6 Sol |
|---|---|---|
| APEX-Agents | 43% | 51.4% |
| OSWorld 2.0 | 9.3% | 27.3% |
| GBAEval | 48.8% | 52.6% |
| GDP.pdf | 18% | 30.7% |
| LMArena Search | 1221 | 1257 |
| Vending-Bench 2 | 7,204 | 9,619 |
| Terminal-Bench | 53.4% | — |
| τ²-bench Banking | — | 46.9% |
| DeepResearch Bench | 54.9% | — |
| OSWorld | 72.1% | — |
| PostTrainBench | — | 36.2% |
| BALROG | — | 60% |
| ExploitBench | 23.6% | — |
Reasoning GPT-5.6 Sol leads
Claude Sonnet 4.6: 46.1 (#40), GPT-5.6 Sol: 74.7 (#6)
| Benchmark | Claude Sonnet 4.6 | GPT-5.6 Sol |
|---|---|---|
| ARC-AGI-2 | 60.4% | 92.5% |
| NYT Connections (extended) | 80.9% | 93.8% |
| ARC-AGI-1 | 86.5% | 97.5% |
| CritPt | 3.1% | 32.3% |
| Chess Puzzles | 13% | 64% |
| LMArena Hard Prompts | 1485 | 1481 |
| Mystery Game Puzzles | 16% | 58% |
| DTBench | 89.9% | 96% |
| LMCA | 46.5% | 59.2% |
| Epoch Capabilities Index | 152.25 | 161.99 |
| SimpleBench | — | 71.7% |
| Kagi LLM Benchmark | — | 67% |
| EnigmaEval | — | 37.1% |
| Thematic Generalization | 76.3% | — |
| EBR-Bench | — | 44.8% |
| Surface Evolver Bench | — | 93.1% |
| Bench to the Future 3 | — | 0.14 |
| ForecastBench | 62 | — |
Math GPT-5.6 Sol leads
Claude Sonnet 4.6: 51.9 (#41), GPT-5.6 Sol: 85.5 (#7)
| Benchmark | Claude Sonnet 4.6 | GPT-5.6 Sol |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 85.8% | 100% |
| ProofBench | 45% | 83% |
| LMArena Math | 1463 | 1472 |
| FrontierMath (Tiers 1-3) | — | 89.1% |
| FrontierMath Tier 4 | — | 82.9% |
| FrontierMath (Feb 2025 set) | 32.4% | — |
| FrontierMath Erdős | — | 0% |
| FrontierMath Tier 4 (v1) | 8.3% | — |
Knowledge GPT-5.6 Sol leads
Claude Sonnet 4.6: 51.6 (#60), GPT-5.6 Sol: 64.3 (#16)
| Benchmark | Claude Sonnet 4.6 | GPT-5.6 Sol |
|---|---|---|
| GPQA Diamond | 87.4% | 93.5% |
| SimpleQA Verified | 35.5% | 69.7% |
| Vectara Hallucination Rate | 10.6% | 12.4% |
| LMArena Expert | 1500 | 1517 |
Multimodal GPT-5.6 Sol leads
Claude Sonnet 4.6: 38.0 (#61), GPT-5.6 Sol: 48.9 (#8)
| Benchmark | Claude Sonnet 4.6 | GPT-5.6 Sol |
|---|---|---|
| LMArena Vision | 1283 | 1281 |
| Blueprint-Bench 2 | 6.7% | 33.6% |
| LMArena Document | 1482 | 1483 |
| Furniture Assembly | — | 56.7% |
Multilingual Too close to call
Claude Sonnet 4.6: 54.6 (#39), GPT-5.6 Sol: 55.4 (#32)
| Benchmark | Claude Sonnet 4.6 | GPT-5.6 Sol |
|---|---|---|
| LMArena Non-English | 1440 | 1450 |
| LMArena Chinese | 1491 | 1523 |
| LMArena French | 1465 | 1475 |
| LMArena German | 1428 | 1476 |
| LMArena Japanese | 1419 | 1475 |
| LMArena Korean | 1411 | 1434 |
| LMArena Russian | 1440 | 1466 |
| LMArena Spanish | 1465 | 1449 |
Instruction Following Too close to call
Claude Sonnet 4.6: 77.6 (#23), GPT-5.6 Sol: 77.8 (#16)
| Benchmark | Claude Sonnet 4.6 | GPT-5.6 Sol |
|---|---|---|
| LMArena Instruction Following | 1476 | 1480 |
Long Context Too close to call
Claude Sonnet 4.6: 45.3 (#40), GPT-5.6 Sol: 45.3 (#42)
| Benchmark | Claude Sonnet 4.6 | GPT-5.6 Sol |
|---|---|---|
| LMArena Longer Query | 1479 | 1479 |
Writing & Preference GPT-5.6 Sol leads
Claude Sonnet 4.6: 70.4 (#22), GPT-5.6 Sol: 73.5 (#12)
| Benchmark | Claude Sonnet 4.6 | GPT-5.6 Sol |
|---|---|---|
| LMArena Text | 1458 | 1456 |
| LMArena Creative Writing | 1435 | 1450 |
| EQ-Bench Creative Writing | 1810 | 1972 |
| EQ-Bench 4 | 1207 | 1250 |
| LMArena Multi-Turn | 1464 | 1456 |
Frequently asked questions
Is Claude Sonnet 4.6 better than GPT-5.6 Sol?
GPT-5.6 Sol is the stronger model overall, scoring 64.4 to 50.1 on the Noometry Index.
Which is cheaper, Claude Sonnet 4.6 or GPT-5.6 Sol?
Claude Sonnet 4.6 is cheaper. It lists at $3 per million input tokens and $15 per million output tokens; GPT-5.6 Sol lists at $4 and $20.
Is Claude Sonnet 4.6 or GPT-5.6 Sol better for coding?
GPT-5.6 Sol scores higher on coding benchmarks: 61.1 versus 46.0 in the Noometry coding category.
Which has the bigger context window?
GPT-5.6 Sol does, with 1.05M tokens against 1M.
How many benchmarks do Claude Sonnet 4.6 and GPT-5.6 Sol share?
48 benchmarks have published results for both models. Claude Sonnet 4.6 has 57 scored results on Noometry and GPT-5.6 Sol has 64.