Model comparison
Claude Fable 5.1 vs GPT-6 Astra: Benchmark Comparison
Claude Fable 5.1 has the higher score on 15 of 23 shared benchmarks; GPT-6 Astra leads on 8.
The largest observed score gap is 33.97 pts on SRE Bench , where GPT-6 Astra leads.
Reported ±1 standard-error ranges overlap on 9 of 21 shared benchmarks with comparable uncertainty data. This is not a pairwise statistical significance test.
Shared benchmark results
Scores and reported standard errors come from the latest visible Vals benchmark version available for both models.
| Benchmark | Claude Fable 5.1 | GPT-6 Astra | Gap | Reported uncertainty |
|---|---|---|---|---|
| Vals Index | 65.83% ±1.11 | 63.13% ±1.19 | 2.70 pts | Reported ±1 SE ranges do not overlap |
| Vals RSI Index | 36.09% | 27.07% | 9.02 pts | Uncertainty comparison unavailable |
| Harvey's Legal Agent Benchmark | 6.67% ±1.17 | 5.42% ±1.17 | 1.25 pts | Reported ±1 SE ranges overlap |
| Legal Research Bench | 55.29% ±3.46 | 39.42% ±3.40 | 15.86 pts | Reported ±1 SE ranges do not overlap |
| EMB | 76.67% ±2.08 | 71.70% ±2.48 | 4.97 pts | Reported ±1 SE ranges do not overlap |
| Finance Agent (v2) | 58.88% ±2.06 | 53.54% ±2.08 | 5.34 pts | Reported ±1 SE ranges do not overlap |
| Tax Agent Bench | 77.64% ±2.83 | 63.34% ±3.15 | 14.31 pts | Reported ±1 SE ranges do not overlap |
| MedCode | 53.51% ±2.17 | 48.49% ±2.13 | 5.02 pts | Reported ±1 SE ranges do not overlap |
| MedScribe | 91.29% ±1.95 | 87.91% ±1.94 | 3.39 pts | Reported ±1 SE ranges overlap |
| ProofBench v1.1 | 100.00% ±0.00 | 99.00% ±1.00 | 1.00 pts | Reported ±1 SE ranges overlap |
| MysteryMechanism | 47.75% ±3.36 | 53.15% ±3.36 | 5.41 pts | Reported ±1 SE ranges overlap |
| Terminal-Bench Science | 40.00% ±5.90 | 62.86% ±5.82 | 22.86 pts | Reported ±1 SE ranges do not overlap |
| SAGE | 48.53% ±3.34 | 46.37% ±3.44 | 2.16 pts | Reported ±1 SE ranges overlap |
| Code Migration | 54.61% ±4.81 | 67.74% ±4.22 | 13.13 pts | Reported ±1 SE ranges do not overlap |
| IOI | 90.78% ±4.65 | 100.00% ±0.00 | 9.22 pts | Reported ±1 SE ranges do not overlap |
| ProgramBench | 7.00% ±1.81 | 5.50% ±1.62 | 1.50 pts | Reported ±1 SE ranges overlap |
| Terminal-Bench 4.0 | 58.08% ±3.31 | 59.60% ±4.40 | 1.51 pts | Reported ±1 SE ranges overlap |
| Vibe Code Bench 1-100 | 28.00% ±4.49 | 27.64% ±4.08 | 0.36 pts | Reported ±1 SE ranges overlap |
| Vibe Code Bench v1.1 | 90.26% ±1.57 | 89.59% ±2.17 | 0.67 pts | Reported ±1 SE ranges overlap |
| CyberBench v1.1 | 70.42% ±5.43 | 41.07% ±2.56 | 29.34 pts | Reported ±1 SE ranges do not overlap |
| SRE Bench | 22.90% ±2.60 | 56.87% ±3.07 | 33.97 pts | Reported ±1 SE ranges do not overlap |
| CUA-bench | 13.17% | 19.17% | 6.00 pts | Uncertainty comparison unavailable |
| Time Horizon Index: KSP | 63.33% ±0.00 | 90.50% ±0.00 | 27.17 pts | Reported ±1 SE ranges do not overlap |
Performance by category
| Category | Claude Fable 5.1 average | GPT-6 Astra average |
|---|---|---|
| Legal | 30.98% | 22.42% |
| Finance | 71.06% | 62.86% |
| Healthcare | 72.40% | 68.20% |
| Math | 100.00% | 99.00% |
| Science | 43.87% | 58.00% |
| Education | 48.53% | 46.37% |
| Coding | 54.79% | 58.35% |
| Cyber | 46.66% | 48.97% |
Cost and latency
Cost per test appears only where the benchmark reports it for both models. Latency is the measured completion time for that benchmark.
| Benchmark | Claude Fable 5.1 cost | GPT-6 Astra cost | Claude Fable 5.1 latency | GPT-6 Astra latency |
|---|---|---|---|---|
| Vals Index | $28.71 | $18.46 | 1h17m | 27m48s |
| Vals RSI Index | $298.22 | $759.51 | 90h00m | 90h00m |
| Harvey's Legal Agent Benchmark | $46.21 | $26.16 | 1h52m | 24m49s |
| Legal Research Bench | $23.06 | $10.58 | 57m29s | 24m51s |
| EMB | $15.93 | $5.83 | 29m32s | 13m24s |
| Finance Agent (v2) | $8.35 | $6.82 | 18m14s | 11m33s |
| Tax Agent Bench | $13.18 | $5.80 | 28m34s | 15m01s |
| MedCode | N/A | N/A | 3m35s | 102.86s |
| MedScribe | N/A | N/A | 3m08s | 2m34s |
| ProofBench v1.1 | $2.70 | $1.71 | 7m58s | 5m08s |
| MysteryMechanism | $5.63 | $1.56 | 17m41s | 8m09s |
| Terminal-Bench Science | $38.01 | $20.80 | 2h36m | 1h49m |
| SAGE | N/A | N/A | 105.75s | 41.63s |
| Code Migration | $70.97 | $44.36 | 4h20m | 58m30s |
| IOI | $11.20 | $6.50 | 30m06s | 31m40s |
| ProgramBench | $58.52 | $11.57 | 2h27m | 27m02s |
| Terminal-Bench 4.0 | $17.18 | $9.58 | 1h00m | 35m42s |
| Vibe Code Bench 1-100 | $149.66 | $84.72 | 2h53m | 3h26m |
| Vibe Code Bench v1.1 | $33.37 | $38.51 | 57m40s | 39m04s |
| CyberBench v1.1 | $4.01 | $1.84 | 18m32s | 4m10s |
| SRE Bench | $32.23 | $10.24 | 2h03m | 17m51s |
| CUA-bench | $323.78 | $1821.37 | 37.44s | 32.09s |
| Time Horizon Index: KSP | $4655.01 | $4206.59 | N/A | N/A |
Results available only for Claude Fable 5.1
- LegalBench
- MortgageTax
- TaxEval v2
- GPQA Diamond
- MMLU Pro
- MMMU Pro
- LiveCodeBench
- SkillsBench
- Public Benefits Bench v1.1
Results available only for GPT-6 Astra
- BioMysteryBench