Compare AI models
Choose any two models to see their strengths, specifications and published benchmark results side by side.
Explore 26 models from 10 companies. Your selection is saved in the page link.
| At a glance |  GPT-6 Astra |  Claude Fable 5.1 |
|---|---|---|
| Company | OpenAI | Anthropic |
| Position | Flagship | Flagship |
| Best suited for | The most demanding reasoning and end-to-end work | Demanding reasoning and long-horizon tasks |
| Input context | 1,050,000 tokens | 1,000,000 tokens |
| Maximum output | 128,000 tokens | 128,000 tokens |
| API input / 1M tokens | $10 | $10 |
| API output / 1M tokens | $50 | $50 |
Benchmark results
Percent score for the evaluated model run. The stronger result in each row is shaded.
| Benchmark | GPT-6 Astra | Claude Fable 5.1 |
|---|---|---|
| Epoch AI Benchmarking Hub · Knowledge | ||
| SimpleQA VerifiedFactual questions | 75.6%gpt-6-astra · max | 70.8%claude-fable-5-1 · max |
| Epoch AI Benchmarking Hub · Mathematics | ||
| OTIS AIMECompetition mathematics | 100.0%gpt-6-astra · max | 100.0%claude-fable-5-1 · max |
| FrontierMath Tiers 1–3Advanced mathematics | 93.7%gpt-6-astra · max | 90.2%claude-fable-5-1 · max |
| FrontierMath Tier 4Research-level mathematics | 97.6%gpt-6-astra · max | 87.8%claude-fable-5-1 · max |
| FrontierMath ErdősOpen-ended mathematics | 2.9%gpt-6-astra · max | 0.0%claude-fable-5-1 · max |
| Epoch AI Benchmarking Hub · Reasoning | ||
| EBR BenchExtended reasoning | 76.2%gpt-6-astra · max | 57.1%claude-fable-5-1 · max |
| Chess PuzzlesChess problem solving | 72.0%gpt-6-astra · max | 47.0%claude-fable-5-1 · max |
| Mystery Game PuzzlesDeductive reasoning | 84.0%gpt-6-astra · max | 58.0%claude-fable-5-1 · max |
| Epoch AI Benchmarking Hub · Spatial | ||
| Furniture AssemblySpatial assembly reasoning | 80.0%gpt-6-astra · max | 70.0%claude-fable-5-1 · max |
| LiveBench · Reasoning | ||
| Theory Of Mind | 84.6%gpt-6-astra-max | 80.8%claude-fable-5-1-max-effort |
| Zebra Puzzle | 100.0%gpt-6-astra-max | 100.0%claude-fable-5-1-max-effort |
| Spatial | 98.0%gpt-6-astra-max | 100.0%claude-fable-5-1-max-effort |
| Logic With Navigation | 88.0%gpt-6-astra-max | 86.0%claude-fable-5-1-max-effort |
| LiveBench · Coding | ||
| Code Generation | 80.3%gpt-6-astra-max | 90.1%claude-fable-5-1-max-effort |
| Code Completion | 80.4%gpt-6-astra-max | 82.6%claude-fable-5-1-max-effort |
| LiveBench · Agentic Coding | ||
| JavaScript | 63.6%gpt-6-astra-max | 68.2%claude-fable-5-1-max-effort |
| TypeScript | 43.3%gpt-6-astra-max | 60.0%claude-fable-5-1-max-effort |
| Python | 65.0%gpt-6-astra-max | 70.0%claude-fable-5-1-max-effort |
| LiveBench · Mathematics | ||
| AMPS Hard | 98.0%gpt-6-astra-max | 99.0%claude-fable-5-1-max-effort |
| Integrals With Game | 100.0%gpt-6-astra-max | 99.0%claude-fable-5-1-max-effort |
| Math Comp | 97.1%gpt-6-astra-max | 97.1%claude-fable-5-1-max-effort |
| Olympiad | 92.2%gpt-6-astra-max | 93.0%claude-fable-5-1-max-effort |
| LiveBench · Data Analysis | ||
| Consecutive Events | 90.6%gpt-6-astra-max | 90.6%claude-fable-5-1-max-effort |
| Table Join | 58.4%gpt-6-astra-max | 56.2%claude-fable-5-1-max-effort |
| Table Reformat | 100.0%gpt-6-astra-max | 94.1%claude-fable-5-1-max-effort |
| LiveBench · Language | ||
| Connections | 100.0%gpt-6-astra-max | 99.3%claude-fable-5-1-max-effort |
| Plot Unscrambling | 86.3%gpt-6-astra-max | 75.2%claude-fable-5-1-max-effort |
| Typos | 82.0%gpt-6-astra-max | 94.0%claude-fable-5-1-max-effort |
| LiveBench · Instruction following | ||
| Paraphrase | 78.2%gpt-6-astra-max | 71.5%claude-fable-5-1-max-effort |
| Simplify | 70.5%gpt-6-astra-max | 70.2%claude-fable-5-1-max-effort |
| Story Generation | 74.0%gpt-6-astra-max | 78.1%claude-fable-5-1-max-effort |
| Summarize | 79.5%gpt-6-astra-max | 72.2%claude-fable-5-1-max-effort |
Sources: Epoch AI Benchmarking Hub (CC BY 4.0 (Epoch-run rows)) · LiveBench (2026-06-25 question set, Apache 2.0). Scores are rounded to one decimal; task labels are formatted for readability.
Keep reading
Explore AI →
GPT-6 Astra vs Claude Fable 5.1: Flagship Models Compared
GPT-6 Astra and Claude Fable 5.1 share a premium API price. Compare their published math and factual benchmark scores, context limits, and best uses.
GPT-6 Astra vs Sol vs Luna: Which Should You Use?
A practical guide to OpenAI’s GPT-6 family: when Astra’s capability, Sol’s balance, or Luna’s low cost is the right choice, with a direct API price comparison.
Gemini 3.8 Flash: What’s New, What It Costs, and Who It’s For
Google’s September 2026 Flash model brings a million-token input window, multimodal inputs, Search grounding, and lower-cost Batch and Flex options. Here is the practical breakdown.