Claude Fable 5.1
Claude model for demanding reasoning and long-horizon agentic work
Shogo Agent Index
Supported#10 of 9268.2range 62.7–73.7 · 36% of weight measured
Agentic coding
80
37% measured
Long-horizon autonomy
71
50% measured
Knowledge-work agents
55
52% measured
Computer use
62
estimated
Security & context learning
61
estimated
Agentic benchmarks
| Benchmark | Score | Normalised | Index weight | Harness / source |
|---|---|---|---|---|
| SWE-bench Verified | est. | 8% | predicted 88 | |
| Terminal-Bench 2.0 | est. | 8% | predicted 95 | |
| DeepSWE | est. | 6% | predicted 79 | |
| FrontierSWE | 56.3% | 5% | proximus · $138.55/task | |
| CursorBench | 51.8% | 4% | source · $17.28/task | |
| GSO-Bench | 88.2% | 4% | OpenHands | |
| METR Time Horizons | est. | 10% | predicted 86 | |
| Vending-Bench 2 | $5,422 | 6% | source | |
| PostTrainBench | 44.6% | 4% | Claude Code | |
| APEX-Agents | 68.6% | 8% | source | |
| GDPval | est. | 7% | predicted 69 | |
| Remote Labor Index | 17.9% | 5% | source | |
| DeepResearch Bench | est. | 5% | predicted 52 | |
| OSWorld 2.0 | est. | 6% | predicted 40 | |
| OSWorld | est. | 4% | predicted 95 | |
| Cybench | est. | 5% | predicted 93 | |
| CL-bench | est. | 5% | predicted 30 |
Other benchmarks
| Benchmark | Category | Score | Source |
|---|---|---|---|
| ARC-AGI | Reasoning | 97.5% | source |
| ARC-AGI-2 | Reasoning | 90.0% | source |
| Chess Puzzles | Reasoning | 48.0% | source |
| CritPt | Reasoning | 31.1% | source |
| DTBench | Knowledge | 97.6% | source |
| FrontierCode | Coding | 50.9% | claude-code |
| FrontierMath Tiers 1-3 | Math | 90.2% | source |
| FrontierMath Tier 4 | Math | 87.8% | source |
| GeoBench | Multimodal | 84.0% | source |
| Humanity's Last Exam | Reasoning | 46.5% | source |
| LMCA | Knowledge | 65.5% | source |
| MirrorCode | Coding | 93.3% | source |
| OTIS Mock AIME | Math | 100.0% | source |
| ProofBench | Math | 100.0% | source |
| SciCode | Coding | 63.1% | source |
| SimpleBench | Reasoning | 86.6% | source |
| SimpleQA Verified | Knowledge | 70.8% | source |
| WebDev Arena | Coding | 1758 | source |
| WeirdML v3 | Coding | 26.0% | claude_code |