Anthropic released Claude Opus 5.5 as the first model in the Claude 5.5 family. It costs $4/$20 per million input/output tokens, cuts cache reads to $0.20, and is over 30% faster than Opus 5; Anthropic reports about 40% lower cost on typical workloads. The company reports 66.4% on Terminal-Bench 4.0 and 54.4% on FrontierCode v1.1, while warning that small benchmark margins are becoming less reliable. METR received ten business days of predeployment API access and described the AI-R&D gain over Fable 5.1 as incremental rather than discontinuous.
Anthropic releases Claude Opus 5.5 with lower agent costs
2026-09-22Verified announcementTRIAL
Anthropic released Claude Opus 5.5 as the first model in the Claude 5.5 family. It costs $4/$20 per million input/output tokens, cuts cache reads to $0.20, and is over 30% faster than Opus 5; Anthropic reports about 40% lower cost on typical workloads. The company reports 66.4% on Terminal-Bench 4.0 and 54.4% on FrontierCode v1.1, while warning that small benchmark margins are becoming less reliable. METR received ten business days of predeployment API access and described the AI-R&D gain over Fable 5.1 as incremental rather than discontinuous.
Availability: Anthropic API, AWS, Google Cloud, Azure, Claude Code and GitHub Copilot
Imported from the existing research archive. Migration preserves the recorded judgment and is not a fresh verification.
What this means for you
For teams using premium coding models, cache economics and steps per task can outweigh the headline token price. Opus 5.5 is worth a controlled routing trial because it combines lower serving cost with strong long-horizon coding results. Keep internal acceptance tests: Anthropic's benchmark settings vary by effort and safeguards, and METR did not evaluate alignment or policy compliance.
Evidence & related material
Technical details & scores
| Model | claude-opus-5-5; 1M context in Claude Code |
|---|---|
| Pricing | $4 input, $20 output, $0.20 cache reads and $5 cache writes per 1M tokens |
| Speed | >30% faster than Opus 5; optional fast mode up to 2.5x at $8/$40 |
| Benchmarks | Terminal-Bench 4.0 66.4%; FrontierCode v1.1 54.4%; CursorBench 4.0 57.8% |
| External Evaluation | METR, unpaid agreement, 10 business days of API access; preliminary AI-R&D assessment |
| Availability | Anthropic API, AWS, Google Cloud, Azure, Claude Code and GitHub Copilot |
Technical impact 4
Engineering value 5
Adoption signal 4
Maturity 4
Verification cost 3
Risk 3