Model comparison
GPT-5.5 vs GPT-5.6 Sol
GPT-5.6 Sol is the stronger model overall, scoring 65.0 to 63.4 on the Noometry Index.
Last verified . 61 shared benchmarks.
Summary
- They share 61 benchmarks with published results for both. GPT-5.5 scores higher in 4 categories and GPT-5.6 Sol in 6 categories; 6 gaps are clear of the uncertainty.
- The widest gap is in coding, where GPT-5.6 Sol leads 65.1 to 58.2.
- The biggest single-benchmark swing is GSO: 40.2% for GPT-5.5 and 76.5% for GPT-5.6 Sol.
- GPT-5.6 Sol is cheaper at $4 / $20 per million input/output tokens, against $5 / $30 for GPT-5.5.
Side by side
| GPT-5.5 | GPT-5.6 Sol | |
|---|---|---|
| Provider | OpenAI | OpenAI |
| Noometry Index | 63.4 | 65.0 |
| Released | 2026-04-23 | 2026-07-09 |
| Weights | Proprietary | Proprietary |
| Context window | 1.05M | 1.05M |
| Max output | 128K | 128K |
| Input $ / M tokens | $5 | $4 |
| Output $ / M tokens | $30 | $20 |
| Results tracked | 71 | 65 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding GPT-5.6 Sol leads
GPT-5.5: 58.2 (#17), GPT-5.6 Sol: 65.1 (#7)
| Benchmark | GPT-5.5 | GPT-5.6 Sol |
|---|---|---|
| DeepSWE | 67% | 72.7% |
| FrontierCode | 43% | 47.5% |
| LMArena WebDev | 1513 | 1618 |
| SciCode | 56.1% | 57.1% |
| GSO | 40.2% | 76.5% |
| WeirdML | 84.9% | 89.4% |
| LMArena Coding | 1494 | 1498 |
| MirrorCode | 10% | 20% |
| ALE-Bench | 1,943 | 2,177 |
| SWE-bench Verified | 80.6% | — |
| CursorBench | — | 41.7% |
| FrontierSWE | — | 32.2% |
Agentic & Tool Use Too close to call
GPT-5.5: 50.7 (#6), GPT-5.6 Sol: 50.3 (#7)
| Benchmark | GPT-5.5 | GPT-5.6 Sol |
|---|---|---|
| APEX-Agents | 55.1% | 51.4% |
| OSWorld 2.0 | 13% | 27.3% |
| τ²-bench Banking | 44.6% | 46.9% |
| PostTrainBench | 27.2% | 36.2% |
| GBAEval | 53.2% | 52.6% |
| GDP.pdf | 26% | 30.7% |
| LMArena Search | 1242 | 1257 |
| Vending-Bench 2 | 7,524 | 9,619 |
| Terminal-Bench | 84.7% | — |
| Remote Labor Index | 6.3% | — |
| DeepResearch Bench | 54% | — |
| BALROG | — | 60% |
| ExploitBench | 47.4% | — |
Reasoning GPT-5.6 Sol leads
GPT-5.5: 72.8 (#11), GPT-5.6 Sol: 74.8 (#8)
| Benchmark | GPT-5.5 | GPT-5.6 Sol |
|---|---|---|
| ARC-AGI-2 | 85% | 92.5% |
| SimpleBench | 69% | 71.7% |
| Kagi LLM Benchmark | 88.8% | 67% |
| NYT Connections (extended) | 96.2% | 93.8% |
| ARC-AGI-1 | 95% | 97.5% |
| CritPt | 27.1% | 32.3% |
| Chess Puzzles | 54% | 64% |
| EBR-Bench | 34.3% | 44.8% |
| LMArena Hard Prompts | 1489 | 1484 |
| Mystery Game Puzzles | 56% | 58% |
| DTBench | 96% | 96% |
| LMCA | 54.3% | 59.2% |
| Surface Evolver Bench | 88.1% | 93.1% |
| Bench to the Future 3 | 0.14 | 0.14 |
| Epoch Capabilities Index | 159.1 | 161.66 |
| EnigmaEval | — | 37.1% |
| ForecastBench | 60.6 | — |
Math GPT-5.6 Sol leads
GPT-5.5: 81.7 (#11), GPT-5.6 Sol: 85.6 (#9)
| Benchmark | GPT-5.5 | GPT-5.6 Sol |
|---|---|---|
| FrontierMath (Tiers 1-3) | 85.3% | 89.1% |
| FrontierMath Tier 4 | 72.5% | 82.9% |
| OTIS Mock AIME 2024-2025 | 100% | 100% |
| ProofBench | 50% | 83% |
| LMArena Math | 1486 | 1474 |
| FrontierMath Erdős | 0% | 0% |
| MathArena Final-Answer Competitions | 94.3% | — |
| FrontierMath (Feb 2025 set) | 51.7% | — |
| FrontierMath Tier 4 (v1) | 35.4% | — |
Knowledge Too close to call
GPT-5.5: 64.4 (#17), GPT-5.6 Sol: 64.3 (#18)
| Benchmark | GPT-5.5 | GPT-5.6 Sol |
|---|---|---|
| GPQA Diamond | 94% | 93.5% |
| SimpleQA Verified | 63% | 69.7% |
| Vectara Hallucination Rate | 9.3% | 12.4% |
| LMArena Expert | 1508 | 1516 |
Multimodal GPT-5.6 Sol leads
GPT-5.5: 46.9 (#12), GPT-5.6 Sol: 48.6 (#9)
| Benchmark | GPT-5.5 | GPT-5.6 Sol |
|---|---|---|
| LMArena Vision | 1297 | 1281 |
| Blueprint-Bench 2 | 36.2% | 33.6% |
| Furniture Assembly | 44.2% | 56.7% |
| LMArena Document | 1486 | 1483 |
Multilingual GPT-5.5 leads
GPT-5.5: 56.4 (#20), GPT-5.6 Sol: 55.3 (#32)
| Benchmark | GPT-5.5 | GPT-5.6 Sol |
|---|---|---|
| LMArena Non-English | 1467 | 1452 |
| LMArena Chinese | 1533 | 1527 |
| LMArena French | 1486 | 1477 |
| LMArena German | 1480 | 1476 |
| LMArena Japanese | 1498 | 1471 |
| LMArena Korean | 1460 | 1442 |
| LMArena Russian | 1473 | 1468 |
| LMArena Spanish | 1468 | 1441 |
Instruction Following Too close to call
GPT-5.5: 77.5 (#18), GPT-5.6 Sol: 77.7 (#16)
| Benchmark | GPT-5.5 | GPT-5.6 Sol |
|---|---|---|
| LMArena Instruction Following | 1479 | 1482 |
Long Context GPT-5.5 leads
GPT-5.5: 48.3 (#12), GPT-5.6 Sol: 45.4 (#42)
| Benchmark | GPT-5.5 | GPT-5.6 Sol |
|---|---|---|
| LMArena Longer Query | 1484 | 1480 |
| CL-bench Life | 22.2% | — |
Writing & Preference Too close to call
GPT-5.5: 72.7 (#13), GPT-5.6 Sol: 73.3 (#12)
| Benchmark | GPT-5.5 | GPT-5.6 Sol |
|---|---|---|
| LMArena Text | 1472 | 1457 |
| LMArena Creative Writing | 1455 | 1448 |
| EQ-Bench Creative Writing | 1844 | 1972 |
| EQ-Bench 4 | 1315 | 1250 |
| LMArena Multi-Turn | 1476 | 1460 |
Frequently asked questions
Is GPT-5.5 better than GPT-5.6 Sol?
GPT-5.6 Sol is the stronger model overall, scoring 65.0 to 63.4 on the Noometry Index.
Which is cheaper, GPT-5.5 or GPT-5.6 Sol?
GPT-5.6 Sol is cheaper. It lists at $4 per million input tokens and $20 per million output tokens; GPT-5.5 lists at $5 and $30.
Is GPT-5.5 or GPT-5.6 Sol better for coding?
GPT-5.6 Sol scores higher on coding benchmarks: 65.1 versus 58.2 in the Noometry coding category.
Which has the bigger context window?
Both accept 1.05M tokens.
How many benchmarks do GPT-5.5 and GPT-5.6 Sol share?
61 benchmarks have published results for both models. GPT-5.5 has 71 scored results on Noometry and GPT-5.6 Sol has 65.