Model comparison
Claude Sonnet 5 vs GPT-5.4
GPT-5.4 is the stronger model overall, scoring 59.5 to 54.5 on the Noometry Index.
Last verified . 46 shared benchmarks.
Summary
- They share 46 benchmarks with published results for both. Claude Sonnet 5 scores higher in 1 category and GPT-5.4 in 9 categories; 9 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where GPT-5.4 leads 61.7 to 48.2.
- The biggest single-benchmark swing is ProofBench: 77% for Claude Sonnet 5 and 56% for GPT-5.4.
- Claude Sonnet 5 is cheaper at $2 / $10 per million input/output tokens, against $2.50 / $15 for GPT-5.4.
- GPT-5.4 accepts more context: 1.05M tokens versus 1M.
Side by side
| Claude Sonnet 5 | GPT-5.4 | |
|---|---|---|
| Provider | Anthropic | OpenAI |
| Noometry Index | 54.5 | 59.5 |
| Released | 2026-06-29 | 2026-03-05 |
| Weights | Proprietary | Proprietary |
| Context window | 1M | 1.05M |
| Max output | 128K | 128K |
| Input $ / M tokens | $2 | $2.50 |
| Output $ / M tokens | $10 | $15 |
| Results tracked | 51 | 68 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Claude Sonnet 5 leads
Claude Sonnet 5: 55.5 (#20), GPT-5.4: 52.8 (#28)
| Benchmark | Claude Sonnet 5 | GPT-5.4 |
|---|---|---|
| DeepSWE | 53.8% | 51.8% |
| LMArena WebDev | 1540 | 1465 |
| SciCode | 53.6% | 56.6% |
| GSO | 37.3% | 31.4% |
| WeirdML | 68.8% | 77.7% |
| LMArena Coding | 1483 | 1497 |
| ALE-Bench | 1,463 | 1,607 |
| SWE-bench Verified | — | 76.9% |
| FrontierCode | 42.7% | — |
| CursorBench | 34.1% | — |
| MirrorCode | — | 15.6% |
| AlgoTune | — | 1.85 |
Agentic & Tool Use GPT-5.4 leads
Claude Sonnet 5: 42.7 (#16), GPT-5.4: 46.5 (#12)
| Benchmark | Claude Sonnet 5 | GPT-5.4 |
|---|---|---|
| APEX-Agents | 54.5% | 52.4% |
| GBAEval | 65.3% | 45.1% |
| LMArena Search | 1194 | 1197 |
| Vending-Bench 2 | 6,378 | 6,144 |
| Terminal-Bench | — | 81.8% |
| τ²-bench Banking | — | 39.4% |
| DeepResearch Bench | — | 35.1% |
| PostTrainBench | — | 19% |
| METR Time Horizons | — | 74.3% |
Reasoning GPT-5.4 leads
Claude Sonnet 5: 48.2 (#37), GPT-5.4: 61.7 (#17)
| Benchmark | Claude Sonnet 5 | GPT-5.4 |
|---|---|---|
| NYT Connections (extended) | 75.1% | 91.3% |
| CritPt | 16.9% | 23.4% |
| Chess Puzzles | 35% | 44% |
| LMArena Hard Prompts | 1461 | 1485 |
| Mystery Game Puzzles | 35% | 37% |
| DTBench | 92.5% | 94.4% |
| LMCA | 50% | 52% |
| Epoch Capabilities Index | 156.34 | 156.92 |
| ForecastBench | 61.1 | 59.5 |
| ARC-AGI-2 | — | 74% |
| SimpleBench | 60.6% | — |
| Kagi LLM Benchmark | — | 63.8% |
| ARC-AGI-1 | — | 93.7% |
| EnigmaEval | — | 16% |
| Thematic Generalization | — | 80% |
| EBR-Bench | — | 25.4% |
| Surface Evolver Bench | 60% | — |
| Bench to the Future 3 | 0.14 | — |
Math GPT-5.4 leads
Claude Sonnet 5: 66.1 (#21), GPT-5.4: 73.3 (#14)
| Benchmark | Claude Sonnet 5 | GPT-5.4 |
|---|---|---|
| FrontierMath (Tiers 1-3) | 65.6% | 78.6% |
| FrontierMath Tier 4 | 29.3% | 49% |
| OTIS Mock AIME 2024-2025 | 94.7% | 97.8% |
| ProofBench | 77% | 56% |
| LMArena Math | 1467 | 1488 |
| MathArena Final-Answer Competitions | — | 83.1% |
| FrontierMath (Feb 2025 set) | — | 47.6% |
| FrontierMath Tier 4 (v1) | — | 27.1% |
Knowledge GPT-5.4 leads
Claude Sonnet 5: 55.7 (#42), GPT-5.4: 65.3 (#13)
| Benchmark | Claude Sonnet 5 | GPT-5.4 |
|---|---|---|
| GPQA Diamond | 90.5% | 93.3% |
| SimpleQA Verified | 33.7% | 45.1% |
| LMArena Expert | 1491 | 1507 |
| Humanity's Last Exam | — | 36.2% |
| Vectara Hallucination Rate | — | 7% |
Multimodal GPT-5.4 leads
Claude Sonnet 5: 42.4 (#26), GPT-5.4: 43.8 (#19)
| Benchmark | Claude Sonnet 5 | GPT-5.4 |
|---|---|---|
| LMArena Vision | 1275 | 1302 |
| Blueprint-Bench 2 | 24.9% | 27.1% |
| LMArena Document | 1466 | 1471 |
| Furniture Assembly | — | 37.5% |
Multilingual GPT-5.4 leads
Claude Sonnet 5: 54.0 (#55), GPT-5.4: 56.4 (#22)
| Benchmark | Claude Sonnet 5 | GPT-5.4 |
|---|---|---|
| LMArena Non-English | 1431 | 1465 |
| LMArena Chinese | 1475 | 1520 |
| LMArena French | 1457 | 1493 |
| LMArena German | 1448 | 1472 |
| LMArena Japanese | 1422 | 1484 |
| LMArena Korean | 1411 | 1449 |
| LMArena Russian | 1452 | 1480 |
| LMArena Spanish | 1436 | 1453 |
Instruction Following Too close to call
Claude Sonnet 5: 76.6 (#39), GPT-5.4: 77.3 (#25)
| Benchmark | Claude Sonnet 5 | GPT-5.4 |
|---|---|---|
| LMArena Instruction Following | 1453 | 1469 |
Long Context GPT-5.4 leads
Claude Sonnet 5: 44.8 (#53), GPT-5.4: 50.3 (#8)
| Benchmark | Claude Sonnet 5 | GPT-5.4 |
|---|---|---|
| LMArena Longer Query | 1464 | 1474 |
| CL-bench | — | 27.9% |
| CL-bench Life | — | 21.7% |
Writing & Preference GPT-5.4 leads
Claude Sonnet 5: 69.4 (#24), GPT-5.4: 72.1 (#18)
| Benchmark | Claude Sonnet 5 | GPT-5.4 |
|---|---|---|
| LMArena Text | 1443 | 1469 |
| LMArena Creative Writing | 1416 | 1439 |
| EQ-Bench Creative Writing | 1794 | 1840 |
| EQ-Bench 4 | 1236 | 1272 |
| LMArena Multi-Turn | 1454 | 1482 |
Frequently asked questions
Is Claude Sonnet 5 better than GPT-5.4?
GPT-5.4 is the stronger model overall, scoring 59.5 to 54.5 on the Noometry Index.
Which is cheaper, Claude Sonnet 5 or GPT-5.4?
Claude Sonnet 5 is cheaper. It lists at $2 per million input tokens and $10 per million output tokens; GPT-5.4 lists at $2.50 and $15.
Is Claude Sonnet 5 or GPT-5.4 better for coding?
Claude Sonnet 5 scores higher on coding benchmarks: 55.5 versus 52.8 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 5 and GPT-5.4 share?
46 benchmarks have published results for both models. Claude Sonnet 5 has 51 scored results on Noometry and GPT-5.4 has 68.