Model comparison
Claude Sonnet 4.6 vs GPT-5.4
GPT-5.4 is the stronger model overall, scoring 59.4 to 50.3 on the Noometry Index.
Last verified . 52 shared benchmarks.
Summary
- They share 52 benchmarks with published results for both. Claude Sonnet 4.6 scores higher in 1 category and GPT-5.4 in 9 categories; 9 gaps are clear of the uncertainty.
- The widest gap is in math, where GPT-5.4 leads 73.5 to 52.9.
- The biggest single-benchmark swing is Chess Puzzles: 13% for Claude Sonnet 4.6 and 44% for GPT-5.4.
- GPT-5.4 is cheaper at $2.50 / $15 per million input/output tokens, against $3 / $15 for Claude Sonnet 4.6.
- GPT-5.4 accepts more context: 1.05M tokens versus 1M.
Side by side
| Claude Sonnet 4.6 | GPT-5.4 | |
|---|---|---|
| Provider | Anthropic | OpenAI |
| Noometry Index | 50.3 | 59.4 |
| Released | 2026-02-17 | 2026-03-05 |
| Weights | Proprietary | Proprietary |
| Context window | 1M | 1.05M |
| Max output | 128K | 128K |
| Input $ / M tokens | $3 | $2.50 |
| Output $ / M tokens | $15 | $15 |
| Results tracked | 57 | 68 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding GPT-5.4 leads
Claude Sonnet 4.6: 46.3 (#67), GPT-5.4: 52.6 (#33)
| Benchmark | Claude Sonnet 4.6 | GPT-5.4 |
|---|---|---|
| SWE-bench Verified | 75.2% | 76.9% |
| DeepSWE | 29.9% | 51.8% |
| LMArena WebDev | 1522 | 1465 |
| SciCode | 46.8% | 56.6% |
| WeirdML | 66.1% | 77.7% |
| LMArena Coding | 1504 | 1497 |
| ALE-Bench | 1,327 | 1,607 |
| FrontierCode | 24.3% | — |
| GSO | — | 31.4% |
| MirrorCode | — | 15.6% |
| AlgoTune | — | 1.85 |
Agentic & Tool Use GPT-5.4 leads
Claude Sonnet 4.6: 39.1 (#28), GPT-5.4: 46.5 (#13)
| Benchmark | Claude Sonnet 4.6 | GPT-5.4 |
|---|---|---|
| Terminal-Bench | 53.4% | 81.8% |
| APEX-Agents | 43% | 52.4% |
| DeepResearch Bench | 54.9% | 35.1% |
| GBAEval | 48.8% | 45.1% |
| LMArena Search | 1221 | 1197 |
| Vending-Bench 2 | 7,204 | 6,144 |
| OSWorld 2.0 | 9.3% | — |
| τ²-bench Banking | — | 39.4% |
| OSWorld | 72.1% | — |
| PostTrainBench | — | 19% |
| ExploitBench | 23.6% | — |
| GDP.pdf | 18% | — |
| METR Time Horizons | — | 74.3% |
Reasoning GPT-5.4 leads
Claude Sonnet 4.6: 46.1 (#45), GPT-5.4: 61.8 (#19)
| Benchmark | Claude Sonnet 4.6 | GPT-5.4 |
|---|---|---|
| ARC-AGI-2 | 60.4% | 74% |
| NYT Connections (extended) | 80.9% | 91.3% |
| ARC-AGI-1 | 86.5% | 93.7% |
| CritPt | 3.1% | 23.4% |
| Chess Puzzles | 13% | 44% |
| Thematic Generalization | 76.3% | 80% |
| LMArena Hard Prompts | 1484 | 1485 |
| Mystery Game Puzzles | 16% | 37% |
| DTBench | 89.9% | 94.4% |
| LMCA | 46.5% | 52% |
| Epoch Capabilities Index | 152.24 | 156.81 |
| ForecastBench | 62 | 59.5 |
| Kagi LLM Benchmark | — | 63.8% |
| EnigmaEval | — | 16% |
| EBR-Bench | — | 25.4% |
Math GPT-5.4 leads
Claude Sonnet 4.6: 52.9 (#49), GPT-5.4: 73.5 (#19)
| Benchmark | Claude Sonnet 4.6 | GPT-5.4 |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 85.8% | 97.8% |
| ProofBench | 45% | 56% |
| LMArena Math | 1462 | 1488 |
| FrontierMath (Feb 2025 set) | 32.4% | 47.6% |
| FrontierMath Tier 4 (v1) | 8.3% | 27.1% |
| FrontierMath (Tiers 1-3) | — | 78.6% |
| FrontierMath Tier 4 | — | 49% |
| MathArena Final-Answer Competitions | — | 83.1% |
Knowledge GPT-5.4 leads
Claude Sonnet 4.6: 51.7 (#65), GPT-5.4: 65.3 (#14)
| Benchmark | Claude Sonnet 4.6 | GPT-5.4 |
|---|---|---|
| GPQA Diamond | 87.4% | 93.3% |
| SimpleQA Verified | 35.5% | 45.1% |
| Vectara Hallucination Rate | 10.6% | 7% |
| LMArena Expert | 1500 | 1507 |
| Humanity's Last Exam | — | 36.2% |
Multimodal GPT-5.4 leads
Claude Sonnet 4.6: 38.0 (#68), GPT-5.4: 43.7 (#20)
| Benchmark | Claude Sonnet 4.6 | GPT-5.4 |
|---|---|---|
| LMArena Vision | 1283 | 1303 |
| Blueprint-Bench 2 | 6.7% | 27.1% |
| LMArena Document | 1482 | 1471 |
| Furniture Assembly | — | 37.5% |
Multilingual GPT-5.4 leads
Claude Sonnet 4.6: 54.4 (#41), GPT-5.4: 56.2 (#23)
| Benchmark | Claude Sonnet 4.6 | GPT-5.4 |
|---|---|---|
| LMArena Non-English | 1440 | 1465 |
| LMArena Chinese | 1491 | 1519 |
| LMArena French | 1465 | 1493 |
| LMArena German | 1428 | 1472 |
| LMArena Japanese | 1420 | 1485 |
| LMArena Korean | 1411 | 1448 |
| LMArena Russian | 1440 | 1480 |
| LMArena Spanish | 1464 | 1454 |
Instruction Following Too close to call
Claude Sonnet 4.6: 77.4 (#25), GPT-5.4: 77.1 (#27)
| Benchmark | Claude Sonnet 4.6 | GPT-5.4 |
|---|---|---|
| LMArena Instruction Following | 1475 | 1469 |
Long Context GPT-5.4 leads
Claude Sonnet 4.6: 45.3 (#44), GPT-5.4: 50.3 (#8)
| Benchmark | Claude Sonnet 4.6 | GPT-5.4 |
|---|---|---|
| LMArena Longer Query | 1479 | 1473 |
| CL-bench | — | 27.9% |
| CL-bench Life | — | 21.7% |
Writing & Preference GPT-5.4 leads
Claude Sonnet 4.6: 70.2 (#22), GPT-5.4: 71.9 (#17)
| Benchmark | Claude Sonnet 4.6 | GPT-5.4 |
|---|---|---|
| LMArena Text | 1458 | 1469 |
| LMArena Creative Writing | 1435 | 1439 |
| EQ-Bench Creative Writing | 1810 | 1840 |
| EQ-Bench 4 | 1207 | 1272 |
| LMArena Multi-Turn | 1464 | 1482 |
Frequently asked questions
Is Claude Sonnet 4.6 better than GPT-5.4?
GPT-5.4 is the stronger model overall, scoring 59.4 to 50.3 on the Noometry Index.
Which is cheaper, Claude Sonnet 4.6 or GPT-5.4?
GPT-5.4 is cheaper. It lists at $2.50 per million input tokens and $15 per million output tokens; Claude Sonnet 4.6 lists at $3 and $15.
Is Claude Sonnet 4.6 or GPT-5.4 better for coding?
GPT-5.4 scores higher on coding benchmarks: 52.6 versus 46.3 in the Noometry coding category.
Which has the bigger context window?
GPT-5.4 does, with 1.05M tokens against 1M.
How many benchmarks do Claude Sonnet 4.6 and GPT-5.4 share?
52 benchmarks have published results for both models. Claude Sonnet 4.6 has 57 scored results on Noometry and GPT-5.4 has 68.