Model comparison
GPT-5.4 vs GPT-5.5
GPT-5.5 is the stronger model overall, scoring 63.4 to 59.4 on the Noometry Index. GPT-5.4 costs 2.0× less per token, which makes it the better buy when GPT-5.5's lead doesn't matter for your workload.
Last verified . 62 shared benchmarks.
Summary
- They share 62 benchmarks with published results for both. GPT-5.4 scores higher in 2 categories and GPT-5.5 in 8 categories; 6 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where GPT-5.5 leads 72.8 to 61.8.
- The biggest single-benchmark swing is Kagi LLM Benchmark: 63.8% for GPT-5.4 and 88.8% for GPT-5.5.
- GPT-5.4 is cheaper at $2.50 / $15 per million input/output tokens, against $5 / $30 for GPT-5.5.
Side by side
| GPT-5.4 | GPT-5.5 | |
|---|---|---|
| Provider | OpenAI | OpenAI |
| Noometry Index | 59.4 | 63.4 |
| Released | 2026-03-05 | 2026-04-23 |
| Weights | Proprietary | Proprietary |
| Context window | 1.05M | 1.05M |
| Max output | 128K | 128K |
| Input $ / M tokens | $2.50 | $5 |
| Output $ / M tokens | $15 | $30 |
| Results tracked | 68 | 71 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding GPT-5.5 leads
GPT-5.4: 52.6 (#33), GPT-5.5: 58.2 (#17)
| Benchmark | GPT-5.4 | GPT-5.5 |
|---|---|---|
| SWE-bench Verified | 76.9% | 80.6% |
| DeepSWE | 51.8% | 67% |
| LMArena WebDev | 1465 | 1513 |
| SciCode | 56.6% | 56.1% |
| GSO | 31.4% | 40.2% |
| WeirdML | 77.7% | 84.9% |
| LMArena Coding | 1497 | 1494 |
| MirrorCode | 15.6% | 10% |
| ALE-Bench | 1,607 | 1,943 |
| FrontierCode | — | 43% |
| AlgoTune | 1.85 | — |
Agentic & Tool Use GPT-5.5 leads
GPT-5.4: 46.5 (#13), GPT-5.5: 50.7 (#6)
| Benchmark | GPT-5.4 | GPT-5.5 |
|---|---|---|
| Terminal-Bench | 81.8% | 84.7% |
| APEX-Agents | 52.4% | 55.1% |
| τ²-bench Banking | 39.4% | 44.6% |
| DeepResearch Bench | 35.1% | 54% |
| PostTrainBench | 19% | 27.2% |
| GBAEval | 45.1% | 53.2% |
| LMArena Search | 1197 | 1242 |
| Vending-Bench 2 | 6,144 | 7,524 |
| OSWorld 2.0 | — | 13% |
| Remote Labor Index | — | 6.3% |
| ExploitBench | — | 47.4% |
| GDP.pdf | — | 26% |
| METR Time Horizons | 74.3% | — |
Reasoning GPT-5.5 leads
GPT-5.4: 61.8 (#19), GPT-5.5: 72.8 (#11)
| Benchmark | GPT-5.4 | GPT-5.5 |
|---|---|---|
| ARC-AGI-2 | 74% | 85% |
| Kagi LLM Benchmark | 63.8% | 88.8% |
| NYT Connections (extended) | 91.3% | 96.2% |
| ARC-AGI-1 | 93.7% | 95% |
| CritPt | 23.4% | 27.1% |
| Chess Puzzles | 44% | 54% |
| EBR-Bench | 25.4% | 34.3% |
| LMArena Hard Prompts | 1485 | 1489 |
| Mystery Game Puzzles | 37% | 56% |
| DTBench | 94.4% | 96% |
| LMCA | 52% | 54.3% |
| Epoch Capabilities Index | 156.81 | 159.1 |
| ForecastBench | 59.5 | 60.6 |
| SimpleBench | — | 69% |
| EnigmaEval | 16% | — |
| Thematic Generalization | 80% | — |
| Surface Evolver Bench | — | 88.1% |
| Bench to the Future 3 | — | 0.14 |
Math GPT-5.5 leads
GPT-5.4: 73.5 (#19), GPT-5.5: 81.7 (#11)
| Benchmark | GPT-5.4 | GPT-5.5 |
|---|---|---|
| FrontierMath (Tiers 1-3) | 78.6% | 85.3% |
| FrontierMath Tier 4 | 49% | 72.5% |
| MathArena Final-Answer Competitions | 83.1% | 94.3% |
| OTIS Mock AIME 2024-2025 | 97.8% | 100% |
| ProofBench | 56% | 50% |
| LMArena Math | 1488 | 1486 |
| FrontierMath (Feb 2025 set) | 47.6% | 51.7% |
| FrontierMath Tier 4 (v1) | 27.1% | 35.4% |
| FrontierMath Erdős | — | 0% |
Knowledge Too close to call
GPT-5.4: 65.3 (#14), GPT-5.5: 64.4 (#17)
| Benchmark | GPT-5.4 | GPT-5.5 |
|---|---|---|
| GPQA Diamond | 93.3% | 94% |
| SimpleQA Verified | 45.1% | 63% |
| Vectara Hallucination Rate | 7% | 9.3% |
| LMArena Expert | 1507 | 1508 |
| Humanity's Last Exam | 36.2% | — |
Multimodal GPT-5.5 leads
GPT-5.4: 43.7 (#20), GPT-5.5: 46.9 (#12)
| Benchmark | GPT-5.4 | GPT-5.5 |
|---|---|---|
| LMArena Vision | 1303 | 1297 |
| Blueprint-Bench 2 | 27.1% | 36.2% |
| Furniture Assembly | 37.5% | 44.2% |
| LMArena Document | 1471 | 1486 |
Multilingual Too close to call
GPT-5.4: 56.2 (#23), GPT-5.5: 56.4 (#20)
| Benchmark | GPT-5.4 | GPT-5.5 |
|---|---|---|
| LMArena Non-English | 1465 | 1467 |
| LMArena Chinese | 1519 | 1533 |
| LMArena French | 1493 | 1486 |
| LMArena German | 1472 | 1480 |
| LMArena Japanese | 1485 | 1498 |
| LMArena Korean | 1448 | 1460 |
| LMArena Russian | 1480 | 1473 |
| LMArena Spanish | 1454 | 1468 |
Instruction Following Too close to call
GPT-5.4: 77.1 (#27), GPT-5.5: 77.5 (#18)
| Benchmark | GPT-5.4 | GPT-5.5 |
|---|---|---|
| LMArena Instruction Following | 1469 | 1479 |
Long Context GPT-5.4 leads
GPT-5.4: 50.3 (#8), GPT-5.5: 48.3 (#12)
| Benchmark | GPT-5.4 | GPT-5.5 |
|---|---|---|
| CL-bench Life | 21.7% | 22.2% |
| LMArena Longer Query | 1473 | 1484 |
| CL-bench | 27.9% | — |
Writing & Preference Too close to call
GPT-5.4: 71.9 (#17), GPT-5.5: 72.7 (#13)
| Benchmark | GPT-5.4 | GPT-5.5 |
|---|---|---|
| LMArena Text | 1469 | 1472 |
| LMArena Creative Writing | 1439 | 1455 |
| EQ-Bench Creative Writing | 1840 | 1844 |
| EQ-Bench 4 | 1272 | 1315 |
| LMArena Multi-Turn | 1482 | 1476 |
Frequently asked questions
Is GPT-5.4 better than GPT-5.5?
GPT-5.5 is the stronger model overall, scoring 63.4 to 59.4 on the Noometry Index. GPT-5.4 costs 2.0× less per token, which makes it the better buy when GPT-5.5's lead doesn't matter for your workload.
Which is cheaper, GPT-5.4 or GPT-5.5?
GPT-5.4 is cheaper. It lists at $2.50 per million input tokens and $15 per million output tokens; GPT-5.5 lists at $5 and $30.
Is GPT-5.4 or GPT-5.5 better for coding?
GPT-5.5 scores higher on coding benchmarks: 58.2 versus 52.6 in the Noometry coding category.
Which has the bigger context window?
Both accept 1.05M tokens.
How many benchmarks do GPT-5.4 and GPT-5.5 share?
62 benchmarks have published results for both models. GPT-5.4 has 68 scored results on Noometry and GPT-5.5 has 71.