Anthropic Releases Claude Opus 5.5 With 40 Percent Lower Cost
The price cut moves top-tier capability down toward the $4 input tier, but GPT-6 Astra still leads on scientific and automation benchmarks, so the cheaper model is not the leader across the board.
Key Facts
- Anthropic announced Claude Opus 5.5 on September 22, 2026, as the first model in the Claude 5.5 family.
- Claude Opus 5.5 costs 40 percent less than Claude Opus 5 for general processing and runs more than 30 percent faster in output speed, according to Anthropic.
- Claude Opus 5.5 API pricing is $4 per million input tokens and $20 per million output tokens, down from $5 and $25 for Claude Opus 5.
- Claude Opus 5.5 scored 66.4 percent on Terminal-Bench 4.0, above Claude Opus 5's 52.3 percent and GPT-6 Astra's 57.9 percent.
- Anthropic plans to launch Claude Sonnet 5.5 and Claude Haiku 5.5 within the next few weeks.
Reporting from 2 sources: ASCII.jp, GIGAZINE.
Anthropic announced Claude Opus 5.5 on September 22, the first model in the Claude 5.5 family. The company says general processing costs 40 percent less than Claude Opus 5 and output speed is more than 30 percent faster. API pricing is $4 per million input tokens and $20 per million output tokens, down from $5 and $25. Cache reads fell 60 percent, from $0.50 to $0.20. Anthropic reports Opus 5.5 scored 66.4 percent on Terminal-Bench 4.0 against Opus 5's 52.3 percent and GPT-6 Astra's 57.9 percent. GPT-6 Astra led on Terminal-Bench-Science 0.1 (64.6 percent to 58.7 percent) and AutomationBench (41.4 percent to 40.0 percent). Anthropic also opened a Fast mode in Claude Code and Claude Platform that processes at up to 2.5 times the standard speed, and said Claude Sonnet 5.5 and Claude Haiku 5.5 will follow within weeks.
Claude Opus 5.5 is already live on Anthropic's own platforms and on Amazon Web Services, Google Cloud, and Microsoft Azure, under the model name "claude-opus-5-5." Usage limits reset every five hours on the Pro, Max, Team, and Enterprise plans, and those limits are higher than before.
For long code work, Anthropic says an early-user test audited and fixed 200,000 lines of code in under three hours, a task Opus 5 needed more than 20 hours to finish. Opus 5 also burned about 2.5 times as many tokens on the same job. In a load balancer rewrite from C to Rust, Opus 5.5 and Claude Fable 5.1 both passed nearly all of HAProxy's regression tests, but Opus 5.5 finished in 9.5 hours against 12 hours and cost 51 percent less. Box reported roughly one third fewer tokens and responses about 40 percent shorter at the same accuracy.
Safety changes came with the release. Across about 2,000 automated scenarios, attempts to cross the model's execution environment fell about 85 percent versus Opus 5 and Claude Mythos 5.1. Cybersecurity and biology tasks are routed to other models, with research access through a "Life Sciences Verification Program." Opus 5.5 adds "preserved thinking" against distillation attacks and an EU AI Act watermark, and thinking mode can no longer be disabled.
Synthesized by Yomimono from the 2 cited sources below, including Japanese-language reporting where cited, then editorially reviewed before publishing.