Anthropic launched Claude Opus 5.5 on September 22, 2026, positioning the upgraded frontier model as a faster, cheaper drop-in for the mid-summer Claude Opus 5 without breaking existing agent stacks. The San Francisco lab says the new model runs about 30% faster than Opus 5, costs roughly 20% less per token and matches competitor Fable 5.1's quality on most enterprise workloads.
Efficiency Gains, Not A New Ceiling
Anthropic framed 5.5 as a consolidation release: same reasoning API, better economics. Input tokens now cost $4 per million and output tokens $20 per million, a 20% cut from Opus 5's launch pricing, while prompt-cache reads dropped 60% to $0.20 per million. Pro, Max and Team subscribers received expanded five-hour usage windows and a rate-limit reset on the same day, part of Anthropic's push to keep coding agents inside a single provider rather than routing to cheaper open-source fallbacks.
Coding Benchmarks Lead The Pitch
The company put coding at the center of the launch. Opus 5.5 completed a 200,000-line codebase audit in under three hours in Anthropic's internal FrontierCode harness, versus more than 20 hours for Opus 5, and closed a 680,000-line migration in under a day. The model hit a 39-out-of-40 success rate on Anthropic's web-app latency optimization test suite. Anthropic said 5.5 "beats GPT-6 Astra at roughly 20% of the cost per task" on the same benchmark, echoing recent internal research where Claude now leads 26% of Anthropic's own R&D work.
Same-Day Copilot And Cloud Rollout
Opus 5.5 shipped simultaneously to the Claude apps, the Anthropic API, and Amazon Bedrock, Google Vertex AI and Azure AI Foundry through Anthropic's cloud partners. GitHub also flipped the switch inside Copilot on the same day, exposing 5.5 as a selectable model for chat, agent mode and code review across VS Code, JetBrains, Xcode and github.com. Sonnet 5.5 and Haiku 5.5, the mid-tier and cost-optimized variants, are due "in the coming weeks," Anthropic said.
Why It Matters
The release lands as Anthropic reportedly weighs a pre-IPO model refresh and as rivals push agentic pricing lower. By keeping the top-of-stack model in the same API surface while cutting cost and latency, Anthropic can defend enterprise footholds like Novo Nordisk's Claude Science deployment and hold pricing pressure on OpenAI's Astra and Google's Gemini without introducing new safety or governance work for customers.
Reporting based on coverage from 9to5Mac, GitHub Changelog and Anthropic.