Claude Code sends 33k tokens before reading the prompt; OpenCode sends 7k
Why CEREBRO kept it
Direct token optimization mechanics insight
The text below is an automated extraction of the article at https://systima.ai/blog/claude-code-vs-opencode-token-overhead, stored verbatim in the public cerebro-vault repository. Copyright remains with the original publisher (systima.ai).
Claude Code Is Way More Token-Hungry Than OpenCode. We Measured Exactly How Much We put Claude Code and OpenCode on the same model, the same machine, and the same tasks, then examined everything sent and received. Claude Code is far hungrier: When we asked both harnesses for a one-line reply, Claude Code used roughly 33,000 tokens of system prompt, tool schemas, and injected scaffolding before the prompt even arrived. OpenCode used about 7,000. That first test was on Sonnet 4.5. Re-running on Claude Fable 5 narrowed the gap to about 3.3x, because Claude Code sends newer models a much smaller s
Community take
Anthropic optimizes for capability over token cost, unlike competitors under margin pressure—explaining why Claude Code triggers excessive tool calls even on trivial requests.
Backlinks
Appeared in 1 briefing
Related
Shares tags: ai/llm-mechanics
Also from systima.ai
Only signal from systima.ai so far.