Model comparison
GPT-6 Astra vs Kimi K3
GPT-6 Astra is the stronger model overall, scoring 70.8 to 59.5 on the Noometry Index. Kimi K3 costs 3.3× less per token, which makes it the better buy when GPT-6 Astra's lead doesn't matter for your workload.
Last verified . 45 shared benchmarks.
Summary
- They share 45 benchmarks with published results for both. GPT-6 Astra scores higher in 6 categories and Kimi K3 in 4 categories; 10 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where GPT-6 Astra leads 85.1 to 63.0.
- The biggest single-benchmark swing is FrontierMath Tier 4: 97.6% for GPT-6 Astra and 39% for Kimi K3.
- Kimi K3 is cheaper at $3 / $15 per million input/output tokens, against $10 / $50 for GPT-6 Astra.
- GPT-6 Astra accepts more context: 1.05M tokens versus 1.05M.
- Kimi K3 has downloadable open weights; the other is API-only.
Side by side
| GPT-6 Astra | Kimi K3 | |
|---|---|---|
| Provider | OpenAI | Moonshot AI |
| Noometry Index | 70.8 | 59.5 |
| Released | 2026-09-03 | 2026-07-16 |
| Weights | Proprietary | Open |
| Context window | 1.05M | 1.05M |
| Max output | 128K | 1.05M |
| Input $ / M tokens | $10 | $3 |
| Output $ / M tokens | $50 | $15 |
| Results tracked | 56 | 53 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding GPT-6 Astra leads
GPT-6 Astra: 73.7 (#2), Kimi K3: 61.0 (#10)
| Benchmark | GPT-6 Astra | Kimi K3 |
|---|---|---|
| DeepSWE | 74.1% | 68.5% |
| FrontierCode | 53.3% | 44.2% |
| LMArena WebDev | 1786 | 1654 |
| FrontierSWE | 65.5% | 25.9% |
| SciCode | 56.5% | 59.5% |
| WeirdML | 93.6% | 82.6% |
| LMArena Coding | 1487 | 1508 |
| ALE-Bench | 2,951 | 1,524 |
| GSO | 79.4% | — |
| MirrorCode | 46.7% | — |
Agentic & Tool Use GPT-6 Astra leads
GPT-6 Astra: 52.9 (#3), Kimi K3: 41.8 (#20)
| Benchmark | GPT-6 Astra | Kimi K3 |
|---|---|---|
| APEX-Agents | 64.7% | 50.6% |
| GDP.pdf | 34.2% | 19% |
| Vending-Bench 2 | 15,515 | 5,165 |
| Remote Labor Index | 20.8% | — |
| τ²-bench Banking | — | 37.1% |
| PostTrainBench | — | 32% |
| BALROG | 68.3% | — |
| GBAEval | — | 48.3% |
Reasoning GPT-6 Astra leads
GPT-6 Astra: 85.1 (#1), Kimi K3: 63.0 (#17)
| Benchmark | GPT-6 Astra | Kimi K3 |
|---|---|---|
| ARC-AGI-2 | 95% | 60.4% |
| NYT Connections (extended) | 98.1% | 93.6% |
| ARC-AGI-1 | 98.5% | 94.5% |
| CritPt | 31.7% | 23.4% |
| Chess Puzzles | 72% | 39% |
| LMArena Hard Prompts | 1462 | 1496 |
| Mystery Game Puzzles | 84% | 26% |
| DTBench | 97.3% | 91.2% |
| LMCA | 64.4% | 52.7% |
| Epoch Capabilities Index | 166.45 | 157.45 |
| SimpleBench | — | 60.7% |
| EBR-Bench | 76.2% | — |
| Surface Evolver Bench | — | 95% |
| Bench to the Future 3 | 0.14 | — |
| ForecastBench | — | 61.1 |
Math GPT-6 Astra leads
GPT-6 Astra: 93.5 (#2), Kimi K3: 74.2 (#16)
| Benchmark | GPT-6 Astra | Kimi K3 |
|---|---|---|
| FrontierMath (Tiers 1-3) | 93.7% | 72.2% |
| FrontierMath Tier 4 | 97.6% | 39% |
| OTIS Mock AIME 2024-2025 | 100% | 97.2% |
| ProofBench | 99% | 87% |
| LMArena Math | 1465 | 1491 |
| MathArena Final-Answer Competitions | — | 87.8% |
| FrontierMath Erdős | 2.9% | — |
Knowledge GPT-6 Astra leads
GPT-6 Astra: 75.3 (#1), Kimi K3: 63.2 (#21)
| Benchmark | GPT-6 Astra | Kimi K3 |
|---|---|---|
| GPQA Diamond | 95.8% | 93.1% |
| SimpleQA Verified | 75.6% | 50.6% |
| LMArena Expert | 1483 | 1521 |
| Humanity's Last Exam | 54.8% | — |
| Vectara Hallucination Rate | 8.7% | — |
Multimodal GPT-6 Astra leads
GPT-6 Astra: 55.0 (#3), Kimi K3: 37.8 (#70)
| Benchmark | GPT-6 Astra | Kimi K3 |
|---|---|---|
| Blueprint-Bench 2 | 49.7% | 29.5% |
| Furniture Assembly | 80% | 34.2% |
| LMArena Vision | 1281 | — |
| LMArena Document | 1468 | — |
Multilingual Kimi K3 leads
GPT-6 Astra: 53.7 (#61), Kimi K3: 56.3 (#21)
| Benchmark | GPT-6 Astra | Kimi K3 |
|---|---|---|
| LMArena Non-English | 1430 | 1466 |
| LMArena Chinese | 1484 | 1529 |
| LMArena French | 1456 | 1491 |
| LMArena German | 1440 | 1488 |
| LMArena Japanese | 1379 | 1487 |
| LMArena Korean | 1426 | 1458 |
| LMArena Russian | 1436 | 1482 |
| LMArena Spanish | 1407 | 1472 |
Instruction Following Kimi K3 leads
GPT-6 Astra: 76.3 (#44), Kimi K3: 77.7 (#14)
| Benchmark | GPT-6 Astra | Kimi K3 |
|---|---|---|
| LMArena Instruction Following | 1450 | 1483 |
Long Context Kimi K3 leads
GPT-6 Astra: 44.5 (#62), Kimi K3: 45.8 (#29)
| Benchmark | GPT-6 Astra | Kimi K3 |
|---|---|---|
| LMArena Longer Query | 1456 | 1494 |
Writing & Preference Kimi K3 leads
GPT-6 Astra: 75.3 (#7), Kimi K3: 76.6 (#4)
| Benchmark | GPT-6 Astra | Kimi K3 |
|---|---|---|
| LMArena Text | 1441 | 1476 |
| LMArena Creative Writing | 1418 | 1454 |
| EQ-Bench Creative Writing | 2173 | 2082 |
| LMArena Multi-Turn | 1448 | 1488 |
| EQ-Bench 4 | — | 1339 |
Frequently asked questions
Is GPT-6 Astra better than Kimi K3?
GPT-6 Astra is the stronger model overall, scoring 70.8 to 59.5 on the Noometry Index. Kimi K3 costs 3.3× 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 K3?
Kimi K3 is cheaper. It lists at $3 per million input tokens and $15 per million output tokens; GPT-6 Astra lists at $10 and $50.
Is GPT-6 Astra or Kimi K3 better for coding?
GPT-6 Astra scores higher on coding benchmarks: 73.7 versus 61.0 in the Noometry coding category.
Which has the bigger context window?
GPT-6 Astra does, with 1.05M tokens against 1.05M.
How many benchmarks do GPT-6 Astra and Kimi K3 share?
45 benchmarks have published results for both models. GPT-6 Astra has 56 scored results on Noometry and Kimi K3 has 53.