Anthropic launches Claude Opus 5.5 at $4/$20 with cheaper cache reads

Anthropic released Claude Opus 5.5 on 22 September across Claude for Pro, Max, Team and Enterprise users and through the Claude API as `claude-opus-5-5`. Anthropic also lists availability through Amazon Web Services, Google Cloud and Microsoft Foundry, with zero-data-retention support carried over from earlier Opus releases.

GitHub separately began a gradual Copilot rollout for Pro+, Max, Business and Enterprise users. GitHub lists VS Code, Visual Studio, Copilot CLI, the coding agent, the Copilot app, github.com, mobile, JetBrains IDEs, Xcode and Eclipse as supported surfaces, while Business and Enterprise administrators retain model-policy controls.

Anthropic prices Opus 5.5 at $4 per million input tokens and $20 per million output tokens, 20% below Opus 5 on list token prices. Cache reads fall to $0.20 per million tokens, which Anthropic says is 60% below Opus 5. A faster serving mode is available in Claude Code and on the Claude Platform at $8 per million input tokens and $40 per million output tokens, while US-only inference carries a 1.1x multiplier.

The company says typical token-billed workloads cost about 40% less than with Opus 5 because the model also uses fewer tokens per task. That figure is an Anthropic workload estimate rather than a universal discount. Actual spend will depend on prompt length, reasoning effort, cache reuse, tool calls and whether fast or US-only serving is selected.

Anthropic describes Opus 5.5 as reaching Claude Fable 5.1-level performance on most work and publishes gains across coding, agents and professional tasks. The launch materials also say external evaluators including METR and Frontier Design tested the model before release, and Anthropic says its production safeguards cover cybersecurity, biology and model-distillation risks.

Most headline capability comparisons remain Anthropic or launch-partner measurements, often with model-specific harness and effort settings. They establish how the vendor evaluated the release, but they do not establish an independent cross-provider ranking. GitHub similarly reports encouraging early Copilot tests while presenting them as its own evaluation.

Independent developer Simon Willison reported that Opus 5.5 at max effort exhausted its 128,000-token output budget twice while attempting his SVG pelican test, with each failed run costing $2.56 and taking nearly 20 minutes. His lower-effort runs completed. This is a firsthand observation from one prompt rather than evidence that max effort generally fails.

The result is still a useful operational constraint for long agent sessions: a larger reasoning budget can increase cost and wall-clock time without guaranteeing a completed answer. Reproduction across more tasks and model snapshots is needed before drawing broader conclusions about the max setting, while Anthropic and GitHub directly establish the launch pricing and product availability.