Model comparison
GPT-5.6 Luna vs Kimi K3
Kimi K3 is the stronger model overall, scoring 59.5 to 54.7 on the Noometry Index. GPT-5.6 Luna costs 13× less per token, which makes it the better buy when Kimi K3's lead doesn't matter for your workload.
Last verified . 46 shared benchmarks.
Summary
- They share 46 benchmarks with published results for both. GPT-5.6 Luna scores higher in 2 categories and Kimi K3 in 8 categories; 10 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where Kimi K3 leads 63.0 to 47.5.
- The biggest single-benchmark swing is Surface Evolver Bench: 61.9% for GPT-5.6 Luna and 95% for Kimi K3.
- GPT-5.6 Luna is cheaper at $0.20 / $1.20 per million input/output tokens, against $3 / $15 for Kimi K3.
- GPT-5.6 Luna accepts more context: 1.05M tokens versus 1.05M.
- Kimi K3 has downloadable open weights; the other is API-only.
Side by side
| GPT-5.6 Luna | Kimi K3 | |
|---|---|---|
| Provider | OpenAI | Moonshot AI |
| Noometry Index | 54.7 | 59.5 |
| Released | 2026-07-09 | 2026-07-16 |
| Weights | Proprietary | Open |
| Context window | 1.05M | 1.05M |
| Max output | 128K | 1.05M |
| Input $ / M tokens | $0.20 | $3 |
| Output $ / M tokens | $1.20 | $15 |
| Results tracked | 51 | 53 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Kimi K3 leads
GPT-5.6 Luna: 53.9 (#25), Kimi K3: 60.6 (#8)
| Benchmark | GPT-5.6 Luna | Kimi K3 |
|---|---|---|
| DeepSWE | 67.2% | 68.5% |
| FrontierCode | 39.8% | 44.2% |
| LMArena WebDev | 1517 | 1659 |
| SciCode | 53.6% | 59.5% |
| WeirdML | 60.9% | 82.6% |
| LMArena Coding | 1466 | 1509 |
| ALE-Bench | 1,667 | 1,524 |
| CursorBench | 35.9% | — |
| FrontierSWE | — | 25.9% |
Agentic & Tool Use Kimi K3 leads
GPT-5.6 Luna: 35.3 (#37), Kimi K3: 41.8 (#19)
| Benchmark | GPT-5.6 Luna | Kimi K3 |
|---|---|---|
| GDP.pdf | 22.7% | 19% |
| Vending-Bench 2 | 4,095 | 5,165 |
| APEX-Agents | — | 50.6% |
| τ²-bench Banking | — | 37.1% |
| PostTrainBench | — | 32% |
| BALROG | 45.6% | — |
| GBAEval | — | 48.3% |
Reasoning Kimi K3 leads
GPT-5.6 Luna: 47.5 (#38), Kimi K3: 63.0 (#15)
| Benchmark | GPT-5.6 Luna | Kimi K3 |
|---|---|---|
| ARC-AGI-2 | 59.5% | 60.4% |
| SimpleBench | 46.8% | 60.7% |
| NYT Connections (extended) | 69.4% | 93.6% |
| ARC-AGI-1 | 88% | 94.5% |
| CritPt | 20.6% | 23.4% |
| Chess Puzzles | 40% | 39% |
| LMArena Hard Prompts | 1452 | 1495 |
| Mystery Game Puzzles | 21% | 26% |
| DTBench | 89.1% | 91.2% |
| LMCA | 48.5% | 52.7% |
| Surface Evolver Bench | 61.9% | 95% |
| Epoch Capabilities Index | 156.32 | 157.68 |
| Kagi LLM Benchmark | 49.1% | — |
| ForecastBench | — | 61.1 |
Math GPT-5.6 Luna leads
GPT-5.6 Luna: 77.5 (#12), Kimi K3: 74.0 (#13)
| Benchmark | GPT-5.6 Luna | Kimi K3 |
|---|---|---|
| FrontierMath (Tiers 1-3) | 82.1% | 72.2% |
| FrontierMath Tier 4 | 61% | 39% |
| OTIS Mock AIME 2024-2025 | 98.3% | 97.2% |
| ProofBench | 60% | 87% |
| LMArena Math | 1455 | 1493 |
| MathArena Final-Answer Competitions | — | 87.8% |
Knowledge Kimi K3 leads
GPT-5.6 Luna: 58.6 (#31), Kimi K3: 63.2 (#19)
| Benchmark | GPT-5.6 Luna | Kimi K3 |
|---|---|---|
| GPQA Diamond | 91.6% | 93.1% |
| SimpleQA Verified | 41% | 50.6% |
| LMArena Expert | 1479 | 1520 |
Multimodal GPT-5.6 Luna leads
GPT-5.6 Luna: 43.0 (#22), Kimi K3: 38.4 (#58)
| Benchmark | GPT-5.6 Luna | Kimi K3 |
|---|---|---|
| Blueprint-Bench 2 | 22.6% | 29.5% |
| Furniture Assembly | 42.5% | 34.2% |
| LMArena Vision | 1259 | — |
| LMArena Document | 1457 | — |
Multilingual Kimi K3 leads
GPT-5.6 Luna: 53.0 (#73), Kimi K3: 56.5 (#20)
| Benchmark | GPT-5.6 Luna | Kimi K3 |
|---|---|---|
| LMArena Non-English | 1418 | 1466 |
| LMArena Chinese | 1469 | 1528 |
| LMArena French | 1457 | 1488 |
| LMArena German | 1459 | 1474 |
| LMArena Japanese | 1416 | 1492 |
| LMArena Korean | 1414 | 1465 |
| LMArena Russian | 1428 | 1481 |
| LMArena Spanish | 1445 | 1478 |
Instruction Following Kimi K3 leads
GPT-5.6 Luna: 76.0 (#48), Kimi K3: 77.8 (#13)
| Benchmark | GPT-5.6 Luna | Kimi K3 |
|---|---|---|
| LMArena Instruction Following | 1440 | 1481 |
Long Context Kimi K3 leads
GPT-5.6 Luna: 43.9 (#78), Kimi K3: 45.8 (#27)
| Benchmark | GPT-5.6 Luna | Kimi K3 |
|---|---|---|
| LMArena Longer Query | 1437 | 1494 |
Writing & Preference Kimi K3 leads
GPT-5.6 Luna: 68.3 (#27), Kimi K3: 76.7 (#4)
| Benchmark | GPT-5.6 Luna | Kimi K3 |
|---|---|---|
| LMArena Text | 1432 | 1476 |
| LMArena Creative Writing | 1397 | 1452 |
| EQ-Bench Creative Writing | 1829 | 2082 |
| EQ-Bench 4 | 1156 | 1339 |
| LMArena Multi-Turn | 1433 | 1488 |
Frequently asked questions
Is GPT-5.6 Luna better than Kimi K3?
Kimi K3 is the stronger model overall, scoring 59.5 to 54.7 on the Noometry Index. GPT-5.6 Luna costs 13× less per token, which makes it the better buy when Kimi K3's lead doesn't matter for your workload.
Which is cheaper, GPT-5.6 Luna or Kimi K3?
GPT-5.6 Luna is cheaper. It lists at $0.20 per million input tokens and $1.20 per million output tokens; Kimi K3 lists at $3 and $15.
Is GPT-5.6 Luna or Kimi K3 better for coding?
Kimi K3 scores higher on coding benchmarks: 60.6 versus 53.9 in the Noometry coding category.
Which has the bigger context window?
GPT-5.6 Luna does, with 1.05M tokens against 1.05M.
How many benchmarks do GPT-5.6 Luna and Kimi K3 share?
46 benchmarks have published results for both models. GPT-5.6 Luna has 51 scored results on Noometry and Kimi K3 has 53.