Model comparison
GPT-5.4 vs GPT-6 Sol
GPT-6 Sol is the stronger model overall, scoring 61.8 to 59.4 on the Noometry Index.
Last verified . 43 shared benchmarks.
Summary
- They share 43 benchmarks with published results for both. GPT-5.4 scores higher in 6 categories and GPT-6 Sol in 4 categories; 8 gaps are clear of the uncertainty.
- The widest gap is in math, where GPT-6 Sol leads 87.2 to 73.5.
- The biggest single-benchmark swing is FrontierMath Tier 4: 49% for GPT-5.4 and 90% for GPT-6 Sol.
- GPT-6 Sol is cheaper at $2 / $10 per million input/output tokens, against $2.50 / $15 for GPT-5.4.
Side by side
| GPT-5.4 | GPT-6 Sol | |
|---|---|---|
| Provider | OpenAI | OpenAI |
| Noometry Index | 59.4 | 61.8 |
| Released | 2026-03-05 | 2026-09-22 |
| Weights | Proprietary | Proprietary |
| Context window | 1.05M | 1.05M |
| Max output | 128K | 128K |
| Input $ / M tokens | $2.50 | $2 |
| Output $ / M tokens | $15 | $10 |
| Results tracked | 68 | 45 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding GPT-6 Sol leads
GPT-5.4: 52.6 (#33), GPT-6 Sol: 60.1 (#11)
| Benchmark | GPT-5.4 | GPT-6 Sol |
|---|---|---|
| DeepSWE | 51.8% | 68.8% |
| LMArena WebDev | 1465 | 1688 |
| SciCode | 56.6% | 57.6% |
| LMArena Coding | 1497 | 1447 |
| ALE-Bench | 1,607 | 2,462 |
| SWE-bench Verified | 76.9% | — |
| FrontierCode | — | 49.3% |
| GSO | 31.4% | — |
| WeirdML | 77.7% | — |
| MirrorCode | 15.6% | — |
| AlgoTune | 1.85 | — |
Agentic & Tool Use GPT-5.4 leads
GPT-5.4: 46.5 (#13), GPT-6 Sol: 37.2 (#36)
| Benchmark | GPT-5.4 | GPT-6 Sol |
|---|---|---|
| APEX-Agents | 52.4% | 54.3% |
| Vending-Bench 2 | 6,144 | 14,428 |
| Terminal-Bench | 81.8% | — |
| τ²-bench Banking | 39.4% | — |
| DeepResearch Bench | 35.1% | — |
| PostTrainBench | 19% | — |
| GBAEval | 45.1% | — |
| GDP.pdf | — | 26.4% |
| LMArena Search | 1197 | — |
| METR Time Horizons | 74.3% | — |
Reasoning GPT-6 Sol leads
GPT-5.4: 61.8 (#19), GPT-6 Sol: 74.0 (#9)
| Benchmark | GPT-5.4 | GPT-6 Sol |
|---|---|---|
| ARC-AGI-2 | 74% | 89.6% |
| NYT Connections (extended) | 91.3% | 90.1% |
| ARC-AGI-1 | 93.7% | 95.5% |
| CritPt | 23.4% | 30.9% |
| EBR-Bench | 25.4% | 53.3% |
| LMArena Hard Prompts | 1485 | 1418 |
| Mystery Game Puzzles | 37% | 56% |
| DTBench | 94.4% | 97.3% |
| LMCA | 52% | 59.1% |
| Epoch Capabilities Index | 156.81 | 162.72 |
| Kagi LLM Benchmark | 63.8% | — |
| Chess Puzzles | 44% | — |
| EnigmaEval | 16% | — |
| Thematic Generalization | 80% | — |
| ForecastBench | 59.5 | — |
Math GPT-6 Sol leads
GPT-5.4: 73.5 (#19), GPT-6 Sol: 87.2 (#7)
| Benchmark | GPT-5.4 | GPT-6 Sol |
|---|---|---|
| FrontierMath (Tiers 1-3) | 78.6% | 89.8% |
| FrontierMath Tier 4 | 49% | 90% |
| OTIS Mock AIME 2024-2025 | 97.8% | 100% |
| ProofBench | 56% | 83% |
| LMArena Math | 1488 | 1402 |
| MathArena Final-Answer Competitions | 83.1% | — |
| FrontierMath (Feb 2025 set) | 47.6% | — |
| FrontierMath Tier 4 (v1) | 27.1% | — |
Knowledge Too close to call
GPT-5.4: 65.3 (#14), GPT-6 Sol: 64.8 (#15)
| Benchmark | GPT-5.4 | GPT-6 Sol |
|---|---|---|
| GPQA Diamond | 93.3% | 94.3% |
| SimpleQA Verified | 45.1% | 60.7% |
| Vectara Hallucination Rate | 7% | 6.5% |
| LMArena Expert | 1507 | 1439 |
| Humanity's Last Exam | 36.2% | — |
Multimodal GPT-6 Sol leads
GPT-5.4: 43.7 (#20), GPT-6 Sol: 47.6 (#10)
| Benchmark | GPT-5.4 | GPT-6 Sol |
|---|---|---|
| LMArena Vision | 1303 | 1245 |
| Blueprint-Bench 2 | 27.1% | 36.9% |
| Furniture Assembly | 37.5% | 58.3% |
| LMArena Document | 1471 | — |
Multilingual GPT-5.4 leads
GPT-5.4: 56.2 (#23), GPT-6 Sol: 50.5 (#118)
| Benchmark | GPT-5.4 | GPT-6 Sol |
|---|---|---|
| LMArena Non-English | 1465 | 1385 |
| LMArena Chinese | 1519 | 1405 |
| LMArena French | 1493 | 1410 |
| LMArena German | 1472 | 1390 |
| LMArena Japanese | 1485 | 1385 |
| LMArena Korean | 1448 | 1341 |
| LMArena Russian | 1480 | 1401 |
| LMArena Spanish | 1454 | 1384 |
Instruction Following GPT-5.4 leads
GPT-5.4: 77.1 (#27), GPT-6 Sol: 74.5 (#94)
| Benchmark | GPT-5.4 | GPT-6 Sol |
|---|---|---|
| LMArena Instruction Following | 1469 | 1412 |
Long Context GPT-5.4 leads
GPT-5.4: 50.3 (#8), GPT-6 Sol: 43.1 (#108)
| Benchmark | GPT-5.4 | GPT-6 Sol |
|---|---|---|
| LMArena Longer Query | 1473 | 1411 |
| CL-bench | 27.9% | — |
| CL-bench Life | 21.7% | — |
Writing & Preference Too close to call
GPT-5.4: 71.9 (#17), GPT-6 Sol: 71.9 (#18)
| Benchmark | GPT-5.4 | GPT-6 Sol |
|---|---|---|
| LMArena Text | 1469 | 1395 |
| LMArena Creative Writing | 1439 | 1378 |
| EQ-Bench Creative Writing | 1840 | 2125 |
| LMArena Multi-Turn | 1482 | 1412 |
| EQ-Bench 4 | 1272 | — |
Frequently asked questions
Is GPT-5.4 better than GPT-6 Sol?
GPT-6 Sol is the stronger model overall, scoring 61.8 to 59.4 on the Noometry Index.
Which is cheaper, GPT-5.4 or GPT-6 Sol?
GPT-6 Sol 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 GPT-5.4 or GPT-6 Sol better for coding?
GPT-6 Sol scores higher on coding benchmarks: 60.1 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-6 Sol share?
43 benchmarks have published results for both models. GPT-5.4 has 68 scored results on Noometry and GPT-6 Sol has 45.