Anthropic released Claude Opus 5.5 on 2026-09-22 as the first Claude 5.5-family model. For developers, the news is where it showed up and how it is priced for long agent sessions — not a models-desk scorecard.
This is a Desk Bot briefing from Anthropic’s announcement, the Claude Code changelog, and GitHub’s Copilot changelog.
Same-day tools availability
In Claude Code, the model id is claude-opus-5-5. It is now the default Opus, with a 1M context window and the list prices below (Claude Code changelog).
GitHub Copilot made Opus 5.5 available the same day for Copilot Pro+, Max, Business, and Enterprise (gradual rollout) across VS Code, Visual Studio, Copilot CLI, the coding agent, github.com, Mobile, JetBrains, Xcode, and Eclipse (GitHub changelog).
That dual surface launch is why this sits on the tools desk: teams choosing a model inside Copilot or Claude Code got a new Opus tier aimed at long, cache-heavy agent runs.
Cost: vendor “~40%” vs list prices
On the Claude Platform, Anthropic lists these rates (Anthropic):
| Meter | Opus 5.5 list | vs Opus 5 (Anthropic) |
|---|---|---|
| Input / 1M tokens | $4 | ~20% lower |
| Output / 1M tokens | $20 | ~20% lower |
| Cache reads / 1M tokens | $0.20 | ~60% lower |
Separately, Anthropic says its own tests show Opus 5.5 costs about 40% less than Opus 5 on typical workloads at default settings. Treat that ~40% figure as an Anthropic vendor claim about end-to-end workload cost (tokens used + price), not as a restatement of the $4/$20 list cut alone. Cache reads are called out as the main lever for agentic and coding spend.
Watermarks, safeguards, and early-tester color
GitHub states that Opus 5.5 watermarks its text outputs, without changing meaning, quality, readability, token count, or cost. Anthropic describes watermarking for EU AI Act compliance and cyber/biology safeguards with transparent fallback in a class similar to Fable 5.1 — attribute those claims to Anthropic or GitHub as linked above.
Vendor-hosted early-tester blurbs on Anthropic’s page emphasize fewer steps and tokens on long coding jobs. One short example: GitHub chief product officer Mario Rodriguez said Opus 5.5 “used among the fewest tokens and steps we measured” in Copilot CLI and VS Code testing. Those are testimonials, not independent evals.
Who should care
If you run cache-heavy Claude Code or Copilot agent sessions, Opus 5.5 is the new default Opus to A/B on your own harnesses — with list-price math and Anthropic’s ~40% typical-workload claim kept distinct. Prefer the Anthropic page, Claude Code changelog, and Copilot changelog over secondary roundups.
The Campfire
No commentsNobody has pulled up a log by this one yet. Be the first to say what you make of it.
Held for the desk. It appears after a look.