Model comparison
Claude Opus 4.8 vs GPT-6 Astra
GPT-6 Astra is the stronger model overall, scoring 70.8 to 60.7 on the Noometry Index. Claude Opus 4.8 costs 2.0× less per token, which makes it the better buy when GPT-6 Astra's lead doesn't matter for your workload.
Last verified . 50 shared benchmarks.
Summary
- They share 50 benchmarks with published results for both. Claude Opus 4.8 scores higher in 3 categories and GPT-6 Astra in 7 categories; 9 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where GPT-6 Astra leads 85.1 to 64.7.
- The biggest single-benchmark swing is Mystery Game Puzzles: 36% for Claude Opus 4.8 and 84% for GPT-6 Astra.
- Claude Opus 4.8 is cheaper at $5 / $25 per million input/output tokens, against $10 / $50 for GPT-6 Astra.
- GPT-6 Astra accepts more context: 1.05M tokens versus 1M.
Side by side
| Claude Opus 4.8 | GPT-6 Astra | |
|---|---|---|
| Provider | Anthropic | OpenAI |
| Noometry Index | 60.7 | 70.8 |
| Released | 2026-05-28 | 2026-09-03 |
| Weights | Proprietary | Proprietary |
| Context window | 1M | 1.05M |
| Max output | 128K | 128K |
| Input $ / M tokens | $5 | $10 |
| Output $ / M tokens | $25 | $50 |
| Results tracked | 65 | 56 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding GPT-6 Astra leads
Claude Opus 4.8: 59.9 (#12), GPT-6 Astra: 73.7 (#2)
| Benchmark | Claude Opus 4.8 | GPT-6 Astra |
|---|---|---|
| DeepSWE | 59% | 74.1% |
| FrontierCode | 46.5% | 53.3% |
| LMArena WebDev | 1556 | 1786 |
| SciCode | 53.5% | 56.5% |
| GSO | 47.1% | 79.4% |
| WeirdML | 82.9% | 93.6% |
| LMArena Coding | 1490 | 1487 |
| ALE-Bench | 1,564 | 2,951 |
| FrontierSWE | — | 65.5% |
| MirrorCode | — | 46.7% |
Agentic & Tool Use GPT-6 Astra leads
Claude Opus 4.8: 47.6 (#11), GPT-6 Astra: 52.9 (#3)
| Benchmark | Claude Opus 4.8 | GPT-6 Astra |
|---|---|---|
| APEX-Agents | 48.9% | 64.7% |
| Remote Labor Index | 8.3% | 20.8% |
| GDP.pdf | 24% | 34.2% |
| Vending-Bench 2 | 5,787 | 15,515 |
| OSWorld 2.0 | 20.6% | — |
| τ²-bench Banking | 39.7% | — |
| DeepResearch Bench | 50.2% | — |
| PostTrainBench | 33.8% | — |
| BALROG | — | 68.3% |
| GBAEval | 70.9% | — |
| LMArena Search | 1204 | — |
Reasoning GPT-6 Astra leads
Claude Opus 4.8: 64.7 (#16), GPT-6 Astra: 85.1 (#1)
| Benchmark | Claude Opus 4.8 | GPT-6 Astra |
|---|---|---|
| ARC-AGI-2 | 72.1% | 95% |
| NYT Connections (extended) | 91.1% | 98.1% |
| ARC-AGI-1 | 92.5% | 98.5% |
| CritPt | 20.9% | 31.7% |
| Chess Puzzles | 34% | 72% |
| EBR-Bench | 28.6% | 76.2% |
| LMArena Hard Prompts | 1482 | 1462 |
| Mystery Game Puzzles | 36% | 84% |
| DTBench | 94.9% | 97.3% |
| LMCA | 57.5% | 64.4% |
| Bench to the Future 3 | 0.14 | 0.14 |
| Epoch Capabilities Index | 158.21 | 166.45 |
| SimpleBench | 64.8% | — |
| Kagi LLM Benchmark | 88.8% | — |
| EnigmaEval | 23.5% | — |
| Surface Evolver Bench | 87.5% | — |
| ForecastBench | 59.9 | — |
Math GPT-6 Astra leads
Claude Opus 4.8: 78.4 (#13), GPT-6 Astra: 93.5 (#2)
| Benchmark | Claude Opus 4.8 | GPT-6 Astra |
|---|---|---|
| FrontierMath (Tiers 1-3) | 80% | 93.7% |
| FrontierMath Tier 4 | 56.1% | 97.6% |
| OTIS Mock AIME 2024-2025 | 98.3% | 100% |
| ProofBench | 69% | 99% |
| LMArena Math | 1487 | 1465 |
| MathArena Final-Answer Competitions | 91.8% | — |
| FrontierMath (Feb 2025 set) | 47.2% | — |
| FrontierMath Erdős | — | 2.9% |
| FrontierMath Tier 4 (v1) | 31.3% | — |
Knowledge GPT-6 Astra leads
Claude Opus 4.8: 61.3 (#29), GPT-6 Astra: 75.3 (#1)
| Benchmark | Claude Opus 4.8 | GPT-6 Astra |
|---|---|---|
| GPQA Diamond | 91% | 95.8% |
| SimpleQA Verified | 53% | 75.6% |
| LMArena Expert | 1502 | 1483 |
| Humanity's Last Exam | — | 54.8% |
| Vectara Hallucination Rate | — | 8.7% |
Multimodal GPT-6 Astra leads
Claude Opus 4.8: 42.9 (#26), GPT-6 Astra: 55.0 (#3)
| Benchmark | Claude Opus 4.8 | GPT-6 Astra |
|---|---|---|
| LMArena Vision | 1294 | 1281 |
| Blueprint-Bench 2 | 14.5% | 49.7% |
| Furniture Assembly | 42.5% | 80% |
| LMArena Document | 1475 | 1468 |
Multilingual Claude Opus 4.8 leads
Claude Opus 4.8: 55.2 (#33), GPT-6 Astra: 53.7 (#61)
| Benchmark | Claude Opus 4.8 | GPT-6 Astra |
|---|---|---|
| LMArena Non-English | 1450 | 1430 |
| LMArena Chinese | 1507 | 1484 |
| LMArena French | 1481 | 1456 |
| LMArena German | 1472 | 1440 |
| LMArena Japanese | 1440 | 1379 |
| LMArena Korean | 1432 | 1426 |
| LMArena Russian | 1474 | 1436 |
| LMArena Spanish | 1466 | 1407 |
Instruction Following Claude Opus 4.8 leads
Claude Opus 4.8: 77.4 (#24), GPT-6 Astra: 76.3 (#44)
| Benchmark | Claude Opus 4.8 | GPT-6 Astra |
|---|---|---|
| LMArena Instruction Following | 1476 | 1450 |
Long Context Too close to call
Claude Opus 4.8: 45.4 (#35), GPT-6 Astra: 44.5 (#62)
| Benchmark | Claude Opus 4.8 | GPT-6 Astra |
|---|---|---|
| LMArena Longer Query | 1483 | 1456 |
Writing & Preference GPT-6 Astra leads
Claude Opus 4.8: 72.0 (#16), GPT-6 Astra: 75.3 (#7)
| Benchmark | Claude Opus 4.8 | GPT-6 Astra |
|---|---|---|
| LMArena Text | 1461 | 1441 |
| LMArena Creative Writing | 1454 | 1418 |
| EQ-Bench Creative Writing | 1840 | 2173 |
| LMArena Multi-Turn | 1476 | 1448 |
| EQ-Bench 4 | 1281 | — |
Frequently asked questions
Is Claude Opus 4.8 better than GPT-6 Astra?
GPT-6 Astra is the stronger model overall, scoring 70.8 to 60.7 on the Noometry Index. Claude Opus 4.8 costs 2.0× less per token, which makes it the better buy when GPT-6 Astra's lead doesn't matter for your workload.
Which is cheaper, Claude Opus 4.8 or GPT-6 Astra?
Claude Opus 4.8 is cheaper. It lists at $5 per million input tokens and $25 per million output tokens; GPT-6 Astra lists at $10 and $50.
Is Claude Opus 4.8 or GPT-6 Astra better for coding?
GPT-6 Astra scores higher on coding benchmarks: 73.7 versus 59.9 in the Noometry coding category.
Which has the bigger context window?
GPT-6 Astra does, with 1.05M tokens against 1M.
How many benchmarks do Claude Opus 4.8 and GPT-6 Astra share?
50 benchmarks have published results for both models. Claude Opus 4.8 has 65 scored results on Noometry and GPT-6 Astra has 56.