Model comparison
GPT-5.5 vs Kimi K3
GPT-5.5 is the stronger model overall, scoring 63.4 to 59.5 on the Noometry Index. Kimi K3 costs 1.9× less per token, which makes it the better buy when GPT-5.5's lead doesn't matter for your workload.
Last verified . 52 shared benchmarks.
Summary
- They share 52 benchmarks with published results for both. GPT-5.5 scores higher in 7 categories and Kimi K3 in 3 categories; 8 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where GPT-5.5 leads 72.8 to 63.0.
- The biggest single-benchmark swing is ProofBench: 50% for GPT-5.5 and 87% for Kimi K3.
- Kimi K3 is cheaper at $3 / $15 per million input/output tokens, against $5 / $30 for GPT-5.5.
- GPT-5.5 accepts more context: 1.05M tokens versus 1.05M.
- Kimi K3 has downloadable open weights; the other is API-only.
Side by side
| GPT-5.5 | Kimi K3 | |
|---|---|---|
| Provider | OpenAI | Moonshot AI |
| Noometry Index | 63.4 | 59.5 |
| Released | 2026-04-23 | 2026-07-16 |
| Weights | Proprietary | Open |
| Context window | 1.05M | 1.05M |
| Max output | 128K | 1.05M |
| Input $ / M tokens | $5 | $3 |
| Output $ / M tokens | $30 | $15 |
| Results tracked | 71 | 53 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Kimi K3 leads
GPT-5.5: 58.2 (#17), Kimi K3: 61.0 (#10)
| Benchmark | GPT-5.5 | Kimi K3 |
|---|---|---|
| DeepSWE | 67% | 68.5% |
| FrontierCode | 43% | 44.2% |
| LMArena WebDev | 1513 | 1654 |
| SciCode | 56.1% | 59.5% |
| WeirdML | 84.9% | 82.6% |
| LMArena Coding | 1494 | 1508 |
| ALE-Bench | 1,943 | 1,524 |
| SWE-bench Verified | 80.6% | — |
| FrontierSWE | — | 25.9% |
| GSO | 40.2% | — |
| MirrorCode | 10% | — |
Agentic & Tool Use GPT-5.5 leads
GPT-5.5: 50.7 (#6), Kimi K3: 41.8 (#20)
| Benchmark | GPT-5.5 | Kimi K3 |
|---|---|---|
| APEX-Agents | 55.1% | 50.6% |
| τ²-bench Banking | 44.6% | 37.1% |
| PostTrainBench | 27.2% | 32% |
| GBAEval | 53.2% | 48.3% |
| GDP.pdf | 26% | 19% |
| Vending-Bench 2 | 7,524 | 5,165 |
| Terminal-Bench | 84.7% | — |
| OSWorld 2.0 | 13% | — |
| Remote Labor Index | 6.3% | — |
| DeepResearch Bench | 54% | — |
| ExploitBench | 47.4% | — |
| LMArena Search | 1242 | — |
Reasoning GPT-5.5 leads
GPT-5.5: 72.8 (#11), Kimi K3: 63.0 (#17)
| Benchmark | GPT-5.5 | Kimi K3 |
|---|---|---|
| ARC-AGI-2 | 85% | 60.4% |
| SimpleBench | 69% | 60.7% |
| NYT Connections (extended) | 96.2% | 93.6% |
| ARC-AGI-1 | 95% | 94.5% |
| CritPt | 27.1% | 23.4% |
| Chess Puzzles | 54% | 39% |
| LMArena Hard Prompts | 1489 | 1496 |
| Mystery Game Puzzles | 56% | 26% |
| DTBench | 96% | 91.2% |
| LMCA | 54.3% | 52.7% |
| Surface Evolver Bench | 88.1% | 95% |
| Epoch Capabilities Index | 159.1 | 157.45 |
| ForecastBench | 60.6 | 61.1 |
| Kagi LLM Benchmark | 88.8% | — |
| EBR-Bench | 34.3% | — |
| Bench to the Future 3 | 0.14 | — |
Math GPT-5.5 leads
GPT-5.5: 81.7 (#11), Kimi K3: 74.2 (#16)
| Benchmark | GPT-5.5 | Kimi K3 |
|---|---|---|
| FrontierMath (Tiers 1-3) | 85.3% | 72.2% |
| FrontierMath Tier 4 | 72.5% | 39% |
| MathArena Final-Answer Competitions | 94.3% | 87.8% |
| OTIS Mock AIME 2024-2025 | 100% | 97.2% |
| ProofBench | 50% | 87% |
| LMArena Math | 1486 | 1491 |
| FrontierMath (Feb 2025 set) | 51.7% | — |
| FrontierMath Erdős | 0% | — |
| FrontierMath Tier 4 (v1) | 35.4% | — |
Knowledge GPT-5.5 leads
GPT-5.5: 64.4 (#17), Kimi K3: 63.2 (#21)
| Benchmark | GPT-5.5 | Kimi K3 |
|---|---|---|
| GPQA Diamond | 94% | 93.1% |
| SimpleQA Verified | 63% | 50.6% |
| LMArena Expert | 1508 | 1521 |
| Vectara Hallucination Rate | 9.3% | — |
Multimodal GPT-5.5 leads
GPT-5.5: 46.9 (#12), Kimi K3: 37.8 (#70)
| Benchmark | GPT-5.5 | Kimi K3 |
|---|---|---|
| Blueprint-Bench 2 | 36.2% | 29.5% |
| Furniture Assembly | 44.2% | 34.2% |
| LMArena Vision | 1297 | — |
| LMArena Document | 1486 | — |
Multilingual Too close to call
GPT-5.5: 56.4 (#20), Kimi K3: 56.3 (#21)
| Benchmark | GPT-5.5 | Kimi K3 |
|---|---|---|
| LMArena Non-English | 1467 | 1466 |
| LMArena Chinese | 1533 | 1529 |
| LMArena French | 1486 | 1491 |
| LMArena German | 1480 | 1488 |
| LMArena Japanese | 1498 | 1487 |
| LMArena Korean | 1460 | 1458 |
| LMArena Russian | 1473 | 1482 |
| LMArena Spanish | 1468 | 1472 |
Instruction Following Too close to call
GPT-5.5: 77.5 (#18), Kimi K3: 77.7 (#14)
| Benchmark | GPT-5.5 | Kimi K3 |
|---|---|---|
| LMArena Instruction Following | 1479 | 1483 |
Long Context GPT-5.5 leads
GPT-5.5: 48.3 (#12), Kimi K3: 45.8 (#29)
| Benchmark | GPT-5.5 | Kimi K3 |
|---|---|---|
| LMArena Longer Query | 1484 | 1494 |
| CL-bench Life | 22.2% | — |
Writing & Preference Kimi K3 leads
GPT-5.5: 72.7 (#13), Kimi K3: 76.6 (#4)
| Benchmark | GPT-5.5 | Kimi K3 |
|---|---|---|
| LMArena Text | 1472 | 1476 |
| LMArena Creative Writing | 1455 | 1454 |
| EQ-Bench Creative Writing | 1844 | 2082 |
| EQ-Bench 4 | 1315 | 1339 |
| LMArena Multi-Turn | 1476 | 1488 |
Frequently asked questions
Is GPT-5.5 better than Kimi K3?
GPT-5.5 is the stronger model overall, scoring 63.4 to 59.5 on the Noometry Index. Kimi K3 costs 1.9× 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-5.5 or Kimi K3?
Kimi K3 is cheaper. It lists at $3 per million input tokens and $15 per million output tokens; GPT-5.5 lists at $5 and $30.
Is GPT-5.5 or Kimi K3 better for coding?
Kimi K3 scores higher on coding benchmarks: 61.0 versus 58.2 in the Noometry coding category.
Which has the bigger context window?
GPT-5.5 does, with 1.05M tokens against 1.05M.
How many benchmarks do GPT-5.5 and Kimi K3 share?
52 benchmarks have published results for both models. GPT-5.5 has 71 scored results on Noometry and Kimi K3 has 53.