Model comparison
Claude Opus 4.8 vs GPT-5.4
Claude Opus 4.8 is the stronger model overall, scoring 60.7 to 59.4 on the Noometry Index. GPT-5.4 costs 1.8× less per token, which makes it the better buy when Claude Opus 4.8's lead doesn't matter for your workload.
Last verified . 58 shared benchmarks.
Summary
- They share 58 benchmarks with published results for both. Claude Opus 4.8 scores higher in 6 categories and GPT-5.4 in 4 categories; 7 gaps are clear of the uncertainty.
- The widest gap is in coding, where Claude Opus 4.8 leads 59.9 to 52.6.
- The biggest single-benchmark swing is GBAEval: 70.9% for Claude Opus 4.8 and 45.1% for GPT-5.4.
- GPT-5.4 is cheaper at $2.50 / $15 per million input/output tokens, against $5 / $25 for Claude Opus 4.8.
- GPT-5.4 accepts more context: 1.05M tokens versus 1M.
Side by side
| Claude Opus 4.8 | GPT-5.4 | |
|---|---|---|
| Provider | Anthropic | OpenAI |
| Noometry Index | 60.7 | 59.4 |
| Released | 2026-05-28 | 2026-03-05 |
| Weights | Proprietary | Proprietary |
| Context window | 1M | 1.05M |
| Max output | 128K | 128K |
| Input $ / M tokens | $5 | $2.50 |
| Output $ / M tokens | $25 | $15 |
| Results tracked | 65 | 68 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Claude Opus 4.8 leads
Claude Opus 4.8: 59.9 (#12), GPT-5.4: 52.6 (#33)
| Benchmark | Claude Opus 4.8 | GPT-5.4 |
|---|---|---|
| DeepSWE | 59% | 51.8% |
| LMArena WebDev | 1556 | 1465 |
| SciCode | 53.5% | 56.6% |
| GSO | 47.1% | 31.4% |
| WeirdML | 82.9% | 77.7% |
| LMArena Coding | 1490 | 1497 |
| ALE-Bench | 1,564 | 1,607 |
| SWE-bench Verified | — | 76.9% |
| FrontierCode | 46.5% | — |
| MirrorCode | — | 15.6% |
| AlgoTune | — | 1.85 |
Agentic & Tool Use Claude Opus 4.8 leads
Claude Opus 4.8: 47.6 (#11), GPT-5.4: 46.5 (#13)
| Benchmark | Claude Opus 4.8 | GPT-5.4 |
|---|---|---|
| APEX-Agents | 48.9% | 52.4% |
| τ²-bench Banking | 39.7% | 39.4% |
| DeepResearch Bench | 50.2% | 35.1% |
| PostTrainBench | 33.8% | 19% |
| GBAEval | 70.9% | 45.1% |
| LMArena Search | 1204 | 1197 |
| Vending-Bench 2 | 5,787 | 6,144 |
| Terminal-Bench | — | 81.8% |
| OSWorld 2.0 | 20.6% | — |
| Remote Labor Index | 8.3% | — |
| GDP.pdf | 24% | — |
| METR Time Horizons | — | 74.3% |
Reasoning Claude Opus 4.8 leads
Claude Opus 4.8: 64.7 (#16), GPT-5.4: 61.8 (#19)
| Benchmark | Claude Opus 4.8 | GPT-5.4 |
|---|---|---|
| ARC-AGI-2 | 72.1% | 74% |
| Kagi LLM Benchmark | 88.8% | 63.8% |
| NYT Connections (extended) | 91.1% | 91.3% |
| ARC-AGI-1 | 92.5% | 93.7% |
| CritPt | 20.9% | 23.4% |
| Chess Puzzles | 34% | 44% |
| EnigmaEval | 23.5% | 16% |
| EBR-Bench | 28.6% | 25.4% |
| LMArena Hard Prompts | 1482 | 1485 |
| Mystery Game Puzzles | 36% | 37% |
| DTBench | 94.9% | 94.4% |
| LMCA | 57.5% | 52% |
| Epoch Capabilities Index | 158.21 | 156.81 |
| ForecastBench | 59.9 | 59.5 |
| SimpleBench | 64.8% | — |
| Thematic Generalization | — | 80% |
| Surface Evolver Bench | 87.5% | — |
| Bench to the Future 3 | 0.14 | — |
Math Claude Opus 4.8 leads
Claude Opus 4.8: 78.4 (#13), GPT-5.4: 73.5 (#19)
| Benchmark | Claude Opus 4.8 | GPT-5.4 |
|---|---|---|
| FrontierMath (Tiers 1-3) | 80% | 78.6% |
| FrontierMath Tier 4 | 56.1% | 49% |
| MathArena Final-Answer Competitions | 91.8% | 83.1% |
| OTIS Mock AIME 2024-2025 | 98.3% | 97.8% |
| ProofBench | 69% | 56% |
| LMArena Math | 1487 | 1488 |
| FrontierMath (Feb 2025 set) | 47.2% | 47.6% |
| FrontierMath Tier 4 (v1) | 31.3% | 27.1% |
Knowledge GPT-5.4 leads
Claude Opus 4.8: 61.3 (#29), GPT-5.4: 65.3 (#14)
| Benchmark | Claude Opus 4.8 | GPT-5.4 |
|---|---|---|
| GPQA Diamond | 91% | 93.3% |
| SimpleQA Verified | 53% | 45.1% |
| LMArena Expert | 1502 | 1507 |
| Humanity's Last Exam | — | 36.2% |
| Vectara Hallucination Rate | — | 7% |
Multimodal Too close to call
Claude Opus 4.8: 42.9 (#26), GPT-5.4: 43.7 (#20)
| Benchmark | Claude Opus 4.8 | GPT-5.4 |
|---|---|---|
| LMArena Vision | 1294 | 1303 |
| Blueprint-Bench 2 | 14.5% | 27.1% |
| Furniture Assembly | 42.5% | 37.5% |
| LMArena Document | 1475 | 1471 |
Multilingual GPT-5.4 leads
Claude Opus 4.8: 55.2 (#33), GPT-5.4: 56.2 (#23)
| Benchmark | Claude Opus 4.8 | GPT-5.4 |
|---|---|---|
| LMArena Non-English | 1450 | 1465 |
| LMArena Chinese | 1507 | 1519 |
| LMArena French | 1481 | 1493 |
| LMArena German | 1472 | 1472 |
| LMArena Japanese | 1440 | 1485 |
| LMArena Korean | 1432 | 1448 |
| LMArena Russian | 1474 | 1480 |
| LMArena Spanish | 1466 | 1454 |
Instruction Following Too close to call
Claude Opus 4.8: 77.4 (#24), GPT-5.4: 77.1 (#27)
| Benchmark | Claude Opus 4.8 | GPT-5.4 |
|---|---|---|
| LMArena Instruction Following | 1476 | 1469 |
Long Context GPT-5.4 leads
Claude Opus 4.8: 45.4 (#35), GPT-5.4: 50.3 (#8)
| Benchmark | Claude Opus 4.8 | GPT-5.4 |
|---|---|---|
| LMArena Longer Query | 1483 | 1473 |
| CL-bench | — | 27.9% |
| CL-bench Life | — | 21.7% |
Writing & Preference Too close to call
Claude Opus 4.8: 72.0 (#16), GPT-5.4: 71.9 (#17)
| Benchmark | Claude Opus 4.8 | GPT-5.4 |
|---|---|---|
| LMArena Text | 1461 | 1469 |
| LMArena Creative Writing | 1454 | 1439 |
| EQ-Bench Creative Writing | 1840 | 1840 |
| EQ-Bench 4 | 1281 | 1272 |
| LMArena Multi-Turn | 1476 | 1482 |
Frequently asked questions
Is Claude Opus 4.8 better than GPT-5.4?
Claude Opus 4.8 is the stronger model overall, scoring 60.7 to 59.4 on the Noometry Index. GPT-5.4 costs 1.8× less per token, which makes it the better buy when Claude Opus 4.8's lead doesn't matter for your workload.
Which is cheaper, Claude Opus 4.8 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 Opus 4.8 lists at $5 and $25.
Is Claude Opus 4.8 or GPT-5.4 better for coding?
Claude Opus 4.8 scores higher on coding benchmarks: 59.9 versus 52.6 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 Opus 4.8 and GPT-5.4 share?
58 benchmarks have published results for both models. Claude Opus 4.8 has 65 scored results on Noometry and GPT-5.4 has 68.