Anthropic’s default recommendation, built for multi-hour autonomous coding, large refactors and other long agent runs. It has a 1M tokens context window. Through Anthropic’s API it costs $4.00 per million input tokens and $20.00 per million output tokens.
Among the tools we track, it’s available in Claude Code, Copilot, Cursor, Kiro, Zed, Amp and Devin. It’s the default Opus model in Claude Code, with fast mode at $8 / $40 per million tokens. Anthropic says it won’t retire before September 22, 2027.
What Anthropic reports
- Anthropic reports 66.4% on Terminal-Bench 4.0 (max effort)
- Anthropic reports 57.8% on CursorBench 4.0 (max effort)
Vendor-reported figures, not independently verified by Code×Radar.
Related changes
- Amp replaces its Dial with a per-thread model selector · November 12, 2026
- Goose 1.59 makes live voice configurable and lets subagents resume · November 10, 2026
- Claude Opus 5.5 arrives and becomes the default on Pro and Team Standard · September 22, 2026
- Claude Opus 5.5 rolls out in GitHub Copilot · September 22, 2026
Sources
- platform.claude.com: overview
- platform.claude.com: pricing
- anthropic.com: claude opus 5 5
- docs.github.com: supported models
- kiro.dev: models
Facts last checked November 15, 2026.