Model comparison
GPT-6 Astra vs Kimi K2.6
GPT-6 Astra is the stronger model overall, scoring 70.3 to 47.5 on the Noometry Index. Kimi K2.6 costs 12× less per token, which makes it the better buy when GPT-6 Astra's lead doesn't matter for your workload.
Last verified . 39 shared benchmarks.
Summary
- They share 39 benchmarks with published results for both. GPT-6 Astra scores higher in 8 categories and Kimi K2.6 in 2 categories; 8 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where GPT-6 Astra leads 85.1 to 39.6.
- The biggest single-benchmark swing is ProofBench: 99% for GPT-6 Astra and 16% for Kimi K2.6.
- Kimi K2.6 is cheaper at $0.95 / $4 per million input/output tokens, against $10 / $50 for GPT-6 Astra.
- GPT-6 Astra accepts more context: 1.05M tokens versus 262K.
- Kimi K2.6 has downloadable open weights; the other is API-only.
Side by side
| GPT-6 Astra | Kimi K2.6 | |
|---|---|---|
| Provider | OpenAI | Moonshot AI |
| Noometry Index | 70.3 | 47.5 |
| Released | 2026-09-03 | 2026-04-20 |
| Weights | Proprietary | Open |
| Context window | 1.05M | 262K |
| Max output | 128K | 262K |
| Input $ / M tokens | $10 | $0.95 |
| Output $ / M tokens | $50 | $4 |
| Results tracked | 51 | 51 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding GPT-6 Astra leads
GPT-6 Astra: 70.5 (#3), Kimi K2.6: 50.5 (#41)
| Benchmark | GPT-6 Astra | Kimi K2.6 |
|---|---|---|
| LMArena WebDev | 1792 | 1509 |
| SciCode | 56.5% | 53.5% |
| WeirdML | 93.6% | 55.9% |
| LMArena Coding | 1488 | 1488 |
| ALE-Bench | 2,951 | 1,093 |
| SWE-bench Verified | — | 76.7% |
| DeepSWE | 74.1% | — |
| FrontierCode | 53.3% | — |
| FrontierSWE | 65.5% | — |
| MirrorCode | 46.7% | — |
Agentic & Tool Use GPT-6 Astra leads
GPT-6 Astra: 52.3 (#3), Kimi K2.6: 21.8 (#128)
| Benchmark | GPT-6 Astra | Kimi K2.6 |
|---|---|---|
| Vending-Bench 2 | 15,515 | 6,205 |
| APEX-Agents | 64.7% | — |
| OSWorld 2.0 | — | 4.6% |
| Remote Labor Index | 20.8% | — |
| BALROG | 68.3% | — |
| ExploitBench | — | 18.4% |
| GBAEval | — | 0.9% |
| GDP.pdf | — | 12% |
Reasoning GPT-6 Astra leads
GPT-6 Astra: 85.1 (#1), Kimi K2.6: 39.6 (#51)
| Benchmark | GPT-6 Astra | Kimi K2.6 |
|---|---|---|
| NYT Connections (extended) | 98.1% | 87.2% |
| CritPt | 31.7% | 8% |
| Chess Puzzles | 72% | 26% |
| EBR-Bench | 76.2% | 2.4% |
| LMArena Hard Prompts | 1464 | 1470 |
| Mystery Game Puzzles | 84% | 18% |
| DTBench | 97.3% | 90.9% |
| LMCA | 64.4% | 37.3% |
| Epoch Capabilities Index | 166.6 | 151.03 |
| ARC-AGI-2 | 95% | — |
| ARC-AGI-1 | 98.5% | — |
| Bench to the Future 3 | 0.14 | — |
Math GPT-6 Astra leads
GPT-6 Astra: 93.4 (#1), Kimi K2.6: 56.8 (#34)
| Benchmark | GPT-6 Astra | Kimi K2.6 |
|---|---|---|
| FrontierMath (Tiers 1-3) | 93.7% | 57.2% |
| FrontierMath Tier 4 | 97.6% | 25.6% |
| OTIS Mock AIME 2024-2025 | 100% | 96.1% |
| ProofBench | 99% | 16% |
| LMArena Math | 1463 | 1475 |
| MathArena Final-Answer Competitions | — | 72.9% |
| FrontierMath (Feb 2025 set) | — | 39% |
| FrontierMath Erdős | 2.9% | — |
| FrontierMath Tier 4 (v1) | — | 14.6% |
Knowledge GPT-6 Astra leads
GPT-6 Astra: 75.3 (#1), Kimi K2.6: 53.9 (#48)
| Benchmark | GPT-6 Astra | Kimi K2.6 |
|---|---|---|
| GPQA Diamond | 95.8% | 90.8% |
| SimpleQA Verified | 75.6% | 34.9% |
| Vectara Hallucination Rate | 8.7% | 10.8% |
| LMArena Expert | 1486 | 1491 |
| Humanity's Last Exam | 54.8% | — |
Multimodal GPT-6 Astra leads
GPT-6 Astra: 55.1 (#3), Kimi K2.6: 31.8 (#95)
| Benchmark | GPT-6 Astra | Kimi K2.6 |
|---|---|---|
| LMArena Vision | 1280 | 1282 |
| Blueprint-Bench 2 | 49.7% | 3.9% |
| Furniture Assembly | 80% | 21.7% |
| LMArena Document | 1468 | 1451 |
Multilingual Kimi K2.6 leads
GPT-6 Astra: 54.0 (#53), Kimi K2.6: 55.1 (#35)
| Benchmark | GPT-6 Astra | Kimi K2.6 |
|---|---|---|
| LMArena Non-English | 1431 | 1446 |
| LMArena Chinese | 1495 | 1521 |
| LMArena French | 1467 | 1471 |
| LMArena Russian | 1434 | 1446 |
| LMArena Spanish | 1412 | 1463 |
| LMArena German | — | 1449 |
| LMArena Japanese | — | 1442 |
| LMArena Korean | — | 1428 |
Instruction Following Too close to call
GPT-6 Astra: 76.6 (#37), Kimi K2.6: 76.5 (#40)
| Benchmark | GPT-6 Astra | Kimi K2.6 |
|---|---|---|
| LMArena Instruction Following | 1454 | 1451 |
Long Context Too close to call
GPT-6 Astra: 44.7 (#57), Kimi K2.6: 44.9 (#50)
| Benchmark | GPT-6 Astra | Kimi K2.6 |
|---|---|---|
| LMArena Longer Query | 1460 | 1468 |
Writing & Preference GPT-6 Astra leads
GPT-6 Astra: 75.7 (#7), Kimi K2.6: 68.7 (#25)
| Benchmark | GPT-6 Astra | Kimi K2.6 |
|---|---|---|
| LMArena Text | 1443 | 1455 |
| LMArena Creative Writing | 1423 | 1434 |
| EQ-Bench Creative Writing | 2173 | 1725 |
| LMArena Multi-Turn | 1457 | 1453 |
| EQ-Bench 4 | — | 1202 |
Frequently asked questions
Is GPT-6 Astra better than Kimi K2.6?
GPT-6 Astra is the stronger model overall, scoring 70.3 to 47.5 on the Noometry Index. Kimi K2.6 costs 12× less per token, which makes it the better buy when GPT-6 Astra's lead doesn't matter for your workload.
Which is cheaper, GPT-6 Astra or Kimi K2.6?
Kimi K2.6 is cheaper. It lists at $0.95 per million input tokens and $4 per million output tokens; GPT-6 Astra lists at $10 and $50.
Is GPT-6 Astra or Kimi K2.6 better for coding?
GPT-6 Astra scores higher on coding benchmarks: 70.5 versus 50.5 in the Noometry coding category.
Which has the bigger context window?
GPT-6 Astra does, with 1.05M tokens against 262K.
How many benchmarks do GPT-6 Astra and Kimi K2.6 share?
39 benchmarks have published results for both models. GPT-6 Astra has 51 scored results on Noometry and Kimi K2.6 has 51.