Claude Opus 5.5 launches at 40% less cost than Opus 5 — matching Fable 5.1-level coding performance
Anthropic's Claude Opus 5.5, released 22 September 2026, costs roughly 40% less to run than Opus 5 and generates output over 30% faster, while matching Claude Fable 5.1 on most agentic coding work — a meaningful shift in what AI-assisted development costs per feature shipped.
26 September 2026
Anthropic launched Claude Opus 5.5 on 22 September 2026, and the headline isn’t a new capability tier — it’s the price-to-performance jump. Opus 5.5 now performs at roughly Fable 5.1’s level on most agentic coding, computer use and multidisciplinary reasoning tasks, but costs around 40% less to run than Opus 5 and produces output more than 30% faster. List pricing dropped too: $4 input / $20 output per million tokens, a 20% cut from Opus 5, with a faster mode at $8/$40 for teams that want up to 2.5x speed on interactive work.
Why the gap between the two numbers matters
The 20% list-price cut and the 40% run-cost drop aren’t the same figure, and the difference is the interesting part. The bigger saving comes from efficiency, not just a lower sticker price — faster generation and better token use on sustained, multi-step coding sessions mean the same feature build now consumes meaningfully less compute than it did on Opus 5, on top of the direct price cut. For AI-assisted development work billed against a fixed budget, that compounds: cheaper per token, and fewer tokens needed to get to the same shipped feature.
Anthropic is also carrying over “preserved thinking” from Fable 5.1 into this launch — a safeguard against third parties extracting a model’s internal reasoning by editing its prior context. That’s a detail more relevant to platform builders than to teams commissioning software, but it signals Anthropic treating Opus 5.5 as flagship infrastructure rather than a stopgap release, which matters if you’re choosing a model family to build a longer engagement around.
So what
If a Claude Fable 5.1-scoped project felt tight against budget a month ago, Opus 5.5 is worth a second look — it’s now landing similar coding output at a lower run cost, which either shrinks the bill or frees up room to scope more into the same budget. The pattern holding through 2026 is consistent: frontier coding models keep getting cheaper and faster roughly every few weeks, and a development partner that isn’t tracking which model is the current best fit for your project is quietly leaving money or scope on the table. See our AI-assisted development approach for how we pick and re-evaluate tooling mid-project, or get in touch if you want a build costed against the current model landscape rather than last quarter’s.