Anthropic released Claude Opus 5.5 today, the first model in its new Claude 5.5 lineup. The company says it performs at the level of Claude Fable 5.1 on most work, while costing 40 percent less to run than Claude Opus 5. The release is also Anthropic’s first since CEO Dario Amodei called for pacing AI development across the industry.
Claude Opus 5.5 Cuts Every Token Price by 20 to 60 Percent
Every part of the pricing sheet moved down for Claude Opus 5‘s successor:
- Input tokens: $4 per million, down from $5
- Output tokens: $20 per million, down from $25
- Cache reads: $0.20 per million, down from $0.50, a 60 percent cut
- Cache writes: $5 per million, down from $6.25
Cache reads make up most of the cost in agentic and coding work, which is why that line dropped furthest. Combined with fewer tokens needed per task, Anthropic estimates a 40 percent overall cost reduction on typical workloads. Claude Opus 5.5 also generates output more than 30 percent faster than Claude Opus 5.
A Fast mode is available in Claude Code and the Claude Platform, running at up to 2.5 times normal speed for $8 per million input tokens and $40 per million output tokens. Anthropic is separately raising five-hour usage limits on Pro, Max, Team, and seat-based Enterprise plans, and giving subscription users a rate limit reset they can save and use whenever they choose.
One Engineer Migrated 680,000 Lines of Code in Under a Day
Anthropic’s early testers reported large jumps on complex work. One completed a 680,000-line code migration in less than a day, a job the company says would have taken an engineering team weeks. In an internal test asking the model to cut load times across every page of a web app, Claude Opus 5.5 succeeded 39 out of 40 times, while Claude Opus 5 made smaller improvements that also changed the app’s behavior.
The gap holds on bigger jobs too, an early tester used Claude Opus 5.5 to audit and fix a 200,000-line codebase in under three hours, a task that took Claude Opus 5 more than 20 hours and 2.5 times the tokens. In a separate test, Anthropic had Claude Opus 5.5 and Claude Fable 5.1 both translate HAProxy from C into Rust. Both rewrites passed nearly all of HAProxy’s own regression tests, but Claude Opus 5.5 finished in 9.5 hours against 12 for Claude Fable 5.1, at 51 percent less cost.
On benchmarks, Claude Opus 5.5 leads the field:
- Terminal-Bench 4.0: 66.4 percent, ahead of Claude Fable 5.1’s 55.8 percent and GPT-6 Astra’s 57.9 percent
- FrontierCode v1.1: 54.4 percent at default effort, beating GPT-6 Astra’s top score of 53.3 percent
- CursorBench 4.0: 57.8 percent, ahead of GPT-5.6 Sol’s 41.7 percent by 11 points
- GDPval-AA v2.1: 1846 Elo, ahead of Claude Fable 5.1’s 1735 and Claude Opus 5’s 1708
Anthropic says Claude Opus 5.5 beats GPT-6 Astra on FrontierCode for about a fifth of the cost per task, and matches GPT-6 Astra on Terminal-Bench 4.0 for about 40 percent of the cost.
Anthropic Links the Launch to Its Own Call for Pacing AI Development
Claude Opus 5.5 posted the best results of any Claude model on Anthropic’s automated behavioral audit, an alignment test covering nearly 2,000 simulated scenarios. Anthropic says the model attempted to cross containment boundaries around 85 percent less often than Claude Opus 5 or Claude Mythos 5.1, and every attempt it made was low severity and self-reported by the model itself.
Because its biology and cybersecurity capabilities are close to Claude Mythos 5.1, Claude Opus 5.5 is expected to ship with the same class of safeguards Anthropic built for Claude Fable 5.1. Most cybersecurity tasks get rerouted to Claude Opus 4.8, and biology research beyond routine use requires an application to Anthropic’s Life Sciences Verification Program. The model also carries preserved thinking, Anthropic’s defense against distillation attacks, for API accounts created on or after August 31, 2026.
Claude Opus 5.5 is available today on Claude.ai, the Claude Platform, and through Amazon Web Services, Google Cloud, and Microsoft Azure, with the model ID claude-opus-5-5 for developers building against the API directly.
Our Take
The 40 percent price cut is impressive, but the more telling one is 85 percent, how much less often Claude Opus 5.5 tried to work around its own boundaries compared to Claude Opus 5 and Claude Mythos 5.1. Anthropic clearly wants this release read as proof that its pacing argument holds up in practice.
For everyday use, the HAProxy and codebase audit numbers are more impactful than benchmarks. A model that finishes the same job in under half the time and at half the tokens changes how teams budget for Claude Code, especially with Claude Sonnet 5.5 and Claude Haiku 5.5 both due in the coming weeks.
