Source caveat
All five scores below come from Anthropic's launch table. OpenAI has not published matching numbers for GPT-5.6 Sol on these evaluations. Treat the Anthropic-side comparison as the only currently available signal, and read it with that caveat in mind.
Benchmark scoreboard
| Benchmark | Fable 5.1 | GPT-5.6 Sol | Delta |
|---|---|---|---|
| Terminal-Bench-Science 0.1 | 52.6% | 22.4% | +30.2 |
| Terminal-Bench 4.0 | 55.8% | 37.3% | +18.5 |
| CursorBench 3.2.0 | 73.4% | 67.2% | +6.2 |
| AutomationBench | 31.4% | 19.6% | +11.8 |
| GDPval-AA v2 | 1853 | 1711 | +142 |
Pricing conflict stated openly
Two prices circulate for GPT-5.6 Sol: llm-stats lists $5 per 1M input and $30 per 1M output. VentureBeat reports a promotional rate of $4 and $20. State both with sources; do not pick the cheaper one to flatter a competitor.
| Source | Input $/1M | Output $/1M | URL |
|---|---|---|---|
| llm-stats | $5 | $30 | https://llm-stats.com/models/gpt-5.6-sol |
| VentureBeat | $4 | $20 (promo) | https://venturebeat.com/... |
Ready to test the workflow?
Create account & add creditsRouting notes
OneInfer does not currently route GPT-5.6 Sol. Use OpenAI direct for the cheapest input rate, or Anthropic direct for the lowest latency in the US east region. The cross-vendor routing decision is about account ownership rather than inference quality.
Frequently asked questions
Who wins between Claude Fable 5.1 and GPT-5.6 Sol?
On the five benchmarks Anthropic published, Claude Fable 5.1 leads on every one, including an 18.5-point gap on Terminal-Bench 4.0 and a 30.2-point gap on Terminal-Bench-Science 0.1. The comparison is one-sided until OpenAI publishes matching scores.
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.