Claude Opus 5.5 costs less than Opus 5 and closes on Fable 5.1

Anthropic released Claude Opus 5.5, its first model since the company said it would pace how fast it pushes the frontier. The headline is not raw capability but the ratio of capability to cost. Anthropic says Opus 5.5 performs at roughly the level of Fable 5.1 on most work while costing about 40 percent less to run on typical workloads, with output generated around 30 percent faster. Input tokens drop to $4 per million and output to $20 per million, both 20 percent cheaper than Opus 5, and cache reads fall 60 percent to $0.20 per million.

The clearest jumps are in coding and agentic tasks. On Terminal-Bench 4.0 the model scores 66.4 percent, up from Opus 5's 52.3 percent, and CursorBench 4.0 rises to 57.8 percent from 46.6. For knowledge work, Anthropic reports 1846 Elo on GDPval-AA v2.1, ahead of both Fable 5.1 and Opus 5. One tester, the company says, completed a 680,000-line code migration in under a day. On computer-use tasks it reaches 81.8 percent partial success on OSWorld 2.0.

Anthropic pairs the release with safety claims: Opus 5.5 scores highest on its automated behavioral audit across 2,000 scenarios, shows an 85 percent reduction in attempts to circumvent containment boundaries versus Opus 5, and runs real-time cybersecurity safeguards that fall back to Opus 4.8 when triggered.

Why it matters

If you run agents or large code migrations, the cost curve just moved: near Fable-level output at 40 percent less means you can either spend the same and do more, or cut your bill. Test it on your own workload before switching, since the benchmark gains are heaviest in coding and may not carry over to your tasks.