Coding sessions are longer and use more context. Claude Opus 5.5 is built with that in mind.
- ID
- 28246
- Status
- summarized
- Published
- 24 Sep 2026, 8:00 AM
- Fetched
- 25 Sep 2026, 4:23 AM
- Provider
- Claude
- Category
- ai-labs
- Original URL
- https://claude.com/blog/claude-opus-5-5-built-for-coding-sessions-that-use-more-context
- Source URL
- https://raw.githubusercontent.com/leontloveless/ai-rss-feeds/main/feeds/claude.xml
Summary
- Score
- 5.5
- Created
- 25 Sep 2026, 4:24 AM
- Tags
- Audience
- developersvibe_codersai_agent_users
What happened
Anthropic says Claude Opus 5.5 costs about 40% less to run than Opus 5 for typical token-billed workloads, by cutting input and output token prices 20% and cached-token reads 60%. The post backs this with Claude Code usage data from March to September 2026: Claude works 3.3x longer per prompt with 40%+ more model calls per prompt, 68% fewer interruptions, developers twice as likely to have a tool server or skill connected, context per request up 2.6x, and the input:output token ratio moving from 189:1 to 324:1.
Why it matters
If you pay for Claude Code by the token, the 60% cached-read cut matters more than the 20% token cut, because the post states cache reads are the majority of agentic and coding work costs. The concrete decision is how you lay out context: keep a stable, cacheable prefix across a long session instead of re-pasting or reordering context each turn. Treat the 40% cost claim and the March-September 2026 usage numbers as vendor-reported, since this is Anthropic's own blog and the excerpt gives no independent measurement.
Discussion angle
Do you actually know your cache-hit rate? Take one long agent session and split its cost into cached reads versus fresh input tokens — then compare that against what you'd save by restructuring the context prefix versus switching models.