Stated negative: cost per benchmark point
On cost per Terminal-Bench 4.0 point, Claude Opus 5 beats Claude Fable 5.1. Fable 5.1 buys the top score and the science capability, not the best value per point. A comparison page that concludes everything favours the newest model is not a comparison page and will not be cited.
Benchmark scoreboard
| Benchmark | Fable 5.1 | Opus 5 | Delta |
|---|---|---|---|
| Terminal-Bench-Science 0.1 | 52.6% | 29.0% | +23.6 |
| Terminal-Bench 4.0 | 55.8% | 52.3% | +3.5 |
| CursorBench 3.2.0 | 73.4% | 70.0% | +3.4 |
| OSWorld 2.0 (partial) | 77.9% | 75.4% | +2.5 |
| OSWorld 2.0 (strict) | 41.7% | 39.6% | +2.1 |
| HLE (no tools) | 60.9% | 56.6% | +4.3 |
| HLE (with tools) | 65.0% | 63.6% | +1.4 |
| AutomationBench | 31.4% | 26.9% | +4.5 |
| GDPval-AA v2 | 1853 | 1824 | +29 |
Price
| Dimension | Fable 5.1 | Opus 5 | Notes |
|---|---|---|---|
| Input / 1M | $10 | $5 | Fable is 2x |
| Output / 1M | $50 | $25 | Fable is 2x |
| Cache read / 1M | $0.25 | not published | Fable wins on long context |
| Cost per TB4.0 point | $0.0358 | $0.0191 | Opus wins |
Ready to test the workflow?
Create account & add creditsWhen to pick Fable 5.1
Agentic scientific work, life-sciences workflows, and Mythos-gated compute where access is available. The 23.6-point lead on Terminal-Bench-Science 0.1 is far outside the standard error band.
When to pick Opus 5
High-volume coding at production scale, where the cost per benchmark point matters more than the peak score, and where the cache hit rate is modest. Opus 5 also runs at half the latency on common workloads.
Frequently asked questions
Is Claude Fable 5.1 better than Claude Opus 5?
Claude Fable 5.1 outscores Claude Opus 5 on every benchmark Anthropic published for both, but by uneven margins. On Terminal-Bench-Science 0.1 the gap is 23.6 points (52.6% against 29.0%). On Terminal-Bench 4.0 it is 3.5 points, which is inside the standard error for the science variant. Claude Opus 5 costs half as much per token and delivers more benchmark points per dollar.
Put Claude Fable 5.1 to work
Fund a controlled evaluation, start with a prepared prompt, and measure quality and cost on your own workload.