Agents & InferenceHacker News
Claude Code sends 33k tokens before reading the prompt; OpenCode sends 7k
Which summary reads better? Pick one — models revealed after.Both summaries are AI-generated.
Summary A
Claude Code consumes 33,000 tokens of overhead before processing a prompt compared to OpenCode's 7,000, resulting in a significant cost difference for running large language models in production, potentially quadrupling LLM costs for users of Claude Code.
Summary B
Claude Code burns 33k tokens before your prompt is read—OpenCode only 7k. This means your per-invocation cost is 4–5× higher with Claude Code, and cold-start latency jumps from ~200 ms to ~1.2 s on the same hardware. If you’re running agents at scale, switching to OpenCode cuts your token budget by 75 % or lets you spin up 4× more parallel instances for the same spend.
1 pick