> ## Documentation Index
> Fetch the complete documentation index at: https://docs.apiyi.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Claude Code: ultra-high cache hit rate, calls as low as \$0.08

> Real-world Claude Opus runs in Claude Code show extremely high cache hit rates — consecutive calls land steadily in the \$0.08–0.12 range (with the 0.95x group discount). Long-context loops and big code reviews keep hitting the cache, making blended cost far lower than straight token math.

**2026/5/3 22:40 (UTC+8)** · Industry News · Anthropic

🚀 **Claude Code: ultra-high cache hit rate, give it a spin** — Real-world runs of `claude-opus-4-7` in Claude Code show extremely high cache hit rates: consecutive calls sit steadily in the \$0.08–0.12 per-call range (already including the `ClaudeCode` group's 0.95x discount), with first-byte latency mostly between 1–4 seconds. Long-context loops, big code reviews, multi-turn conversations all keep landing on the cache — blended cost ends up far lower than straight token math would suggest.

🔐 **Channel sourcing**: Our Claude Code channel runs on direct Anthropic official keys (T4 tier) plus high-quota AWS Bedrock accounts — a professional Claude official-relay API, **not reverse-engineered**. Stability, quotas, and prompt-cache behavior all match the official upstream, so long-running production workloads can plug in with confidence.

<Frame>
  <img src="https://mintcdn.com/apiyillc/_VzXicItDKpC5c0P/images/claude-code-cache-hit-log.png?fit=max&auto=format&n=_VzXicItDKpC5c0P&q=85&s=090af11dce1008cbcf3fac126659474c" alt="Claude Code call log: ~$0.08–0.12 per call, 0.95x discount" width="1642" height="1600" data-path="images/claude-code-cache-hit-log.png" />
</Frame>

📖 Setup guide: [/en/scenarios/programming/claude-code](/en/scenarios/programming/claude-code)

***

← [Back to Live Updates](/en/live) · 📚 [Monthly Archive](/en/live/archive)