Model comparison
GPT-6 Astra vs o3
GPT-6 Astra is the stronger model overall, scoring 70.3 to 47.4 on the Noometry Index. o3 costs 5.7× less per token, which makes it the better buy when GPT-6 Astra's lead doesn't matter for your workload.
Last verified . 31 shared benchmarks.
Summary
- They share 31 benchmarks with published results for both. GPT-6 Astra scores higher in 9 categories and o3 in 1 category; 10 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where GPT-6 Astra leads 85.1 to 32.1.
- The biggest single-benchmark swing is ARC-AGI-2: 95% for GPT-6 Astra and 6.5% for o3.
- o3 is cheaper at $2 / $8 per million input/output tokens, against $10 / $50 for GPT-6 Astra.
- GPT-6 Astra accepts more context: 1.05M tokens versus 200K.
Side by side
| GPT-6 Astra | o3 | |
|---|---|---|
| Provider | OpenAI | OpenAI |
| Noometry Index | 70.3 | 47.4 |
| Released | 2026-09-03 | 2025-04-16 |
| Weights | Proprietary | Proprietary |
| Context window | 1.05M | 200K |
| Max output | 128K | 100K |
| Input $ / M tokens | $10 | $2 |
| Output $ / M tokens | $50 | $8 |
| Results tracked | 51 | 63 |
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), o3: 47.3 (#57)
| Benchmark | GPT-6 Astra | o3 |
|---|---|---|
| WeirdML | 93.6% | 52.4% |
| LMArena Coding | 1488 | 1408 |
| ALE-Bench | 2,951 | 933.55 |
| SWE-bench Verified | — | 62.3% |
| DeepSWE | 74.1% | — |
| FrontierCode | 53.3% | — |
| SWE-bench Verified (bash only) | — | 58.4% |
| Aider Polyglot | — | 81.3% |
| LMArena WebDev | 1792 | — |
| FrontierSWE | 65.5% | — |
| SciCode | 56.5% | — |
| GSO | — | 8.8% |
| MirrorCode | 46.7% | — |
| CadEval | — | 74% |
Agentic & Tool Use GPT-6 Astra leads
GPT-6 Astra: 52.3 (#3), o3: 34.3 (#42)
| Benchmark | GPT-6 Astra | o3 |
|---|---|---|
| APEX-Agents | 64.7% | — |
| Berkeley Function Calling Leaderboard | — | 63% |
| GDPval | — | 30.8% |
| Remote Labor Index | 20.8% | — |
| DeepResearch Bench | — | 45.2% |
| OSWorld | — | 23% |
| BALROG | 68.3% | — |
| LMArena Search | — | 1144 |
| METR Time Horizons | — | 65.4% |
| Vending-Bench 2 | 15,515 | — |
Reasoning GPT-6 Astra leads
GPT-6 Astra: 85.1 (#1), o3: 32.1 (#66)
| Benchmark | GPT-6 Astra | o3 |
|---|---|---|
| ARC-AGI-2 | 95% | 6.5% |
| ARC-AGI-1 | 98.5% | 60.8% |
| CritPt | 31.7% | 1.4% |
| Chess Puzzles | 72% | 38% |
| LMArena Hard Prompts | 1464 | 1402 |
| Mystery Game Puzzles | 84% | 29% |
| DTBench | 97.3% | 84.8% |
| LMCA | 64.4% | 39.7% |
| Epoch Capabilities Index | 166.6 | 146.88 |
| SimpleBench | — | 53.1% |
| Kagi LLM Benchmark | — | 67.6% |
| NYT Connections (extended) | 98.1% | — |
| EnigmaEval | — | 13.1% |
| EBR-Bench | 76.2% | — |
| Bench to the Future 3 | 0.14 | — |
| ForecastBench | — | 62.5 |
Math GPT-6 Astra leads
GPT-6 Astra: 93.4 (#1), o3: 48.9 (#49)
| Benchmark | GPT-6 Astra | o3 |
|---|---|---|
| FrontierMath (Tiers 1-3) | 93.7% | 33.3% |
| OTIS Mock AIME 2024-2025 | 100% | 84.4% |
| LMArena Math | 1463 | 1426 |
| FrontierMath Tier 4 | 97.6% | — |
| ProofBench | 99% | — |
| Omni-MATH | — | 71.4% |
| MATH Level 5 | — | 97.8% |
| FrontierMath (Feb 2025 set) | — | 18.7% |
| FrontierMath Erdős | 2.9% | — |
| FrontierMath Tier 4 (v1) | — | 2.1% |
Knowledge GPT-6 Astra leads
GPT-6 Astra: 75.3 (#1), o3: 54.5 (#47)
| Benchmark | GPT-6 Astra | o3 |
|---|---|---|
| GPQA Diamond | 95.8% | 81.8% |
| Humanity's Last Exam | 54.8% | 20.3% |
| SimpleQA Verified | 75.6% | 49.4% |
| LMArena Expert | 1486 | 1403 |
| MMLU-Pro | — | 85.9% |
| Confabulations | — | 14.4% |
| Vectara Hallucination Rate | 8.7% | — |
| GPQA (HELM) | — | 75.3% |
Multimodal GPT-6 Astra leads
GPT-6 Astra: 55.1 (#3), o3: 40.7 (#35)
| Benchmark | GPT-6 Astra | o3 |
|---|---|---|
| LMArena Vision | 1280 | 1214 |
| GeoBench | — | 74% |
| VPCT | — | 52% |
| Blueprint-Bench 2 | 49.7% | — |
| Furniture Assembly | 80% | — |
| LMArena Document | 1468 | — |
Multilingual GPT-6 Astra leads
GPT-6 Astra: 54.0 (#53), o3: 51.9 (#99)
| Benchmark | GPT-6 Astra | o3 |
|---|---|---|
| LMArena Non-English | 1431 | 1402 |
| LMArena Chinese | 1495 | 1438 |
| LMArena French | 1467 | 1430 |
| LMArena Russian | 1434 | 1407 |
| LMArena Spanish | 1412 | 1395 |
| LMArena German | — | 1420 |
| LMArena Japanese | — | 1404 |
| LMArena Korean | — | 1371 |
Instruction Following GPT-6 Astra leads
GPT-6 Astra: 76.6 (#37), o3: 73.0 (#122)
| Benchmark | GPT-6 Astra | o3 |
|---|---|---|
| LMArena Instruction Following | 1454 | 1368 |
| IFEval | — | 86.9% |
Long Context o3 leads
GPT-6 Astra: 44.7 (#57), o3: 53.2 (#6)
| Benchmark | GPT-6 Astra | o3 |
|---|---|---|
| LMArena Longer Query | 1460 | 1372 |
| Fiction.LiveBench | — | 88.9% |
| CL-bench | — | 17.8% |
Writing & Preference GPT-6 Astra leads
GPT-6 Astra: 75.7 (#7), o3: 63.7 (#62)
| Benchmark | GPT-6 Astra | o3 |
|---|---|---|
| LMArena Text | 1443 | 1410 |
| LMArena Creative Writing | 1423 | 1359 |
| EQ-Bench Creative Writing | 2173 | 1676 |
| LMArena Multi-Turn | 1457 | 1405 |
| Short-Story Creative Writing | — | 83.9% |
| WildBench | — | 86.1% |
Frequently asked questions
Is GPT-6 Astra better than o3?
GPT-6 Astra is the stronger model overall, scoring 70.3 to 47.4 on the Noometry Index. o3 costs 5.7× 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 o3?
o3 is cheaper. It lists at $2 per million input tokens and $8 per million output tokens; GPT-6 Astra lists at $10 and $50.
Is GPT-6 Astra or o3 better for coding?
GPT-6 Astra scores higher on coding benchmarks: 70.5 versus 47.3 in the Noometry coding category.
Which has the bigger context window?
GPT-6 Astra does, with 1.05M tokens against 200K.
How many benchmarks do GPT-6 Astra and o3 share?
31 benchmarks have published results for both models. GPT-6 Astra has 51 scored results on Noometry and o3 has 63.