0
Claude Code uses 33k tokens to read prompt; OpenCode uses 7k tokens
TL;DR: Discussion compares token usage between Claude Code and OpenCode during prompt processing, noting older models and tooling influence token counts. The thread centers on efficiency and cost of token consumption in coding agents, with calls for dynamic context pruning and other token-saving techniques.
A Hacker News thread highlights large token differences between Claude Code and OpenCode when processing prompts. Participants note that older model versions and token-heavy tooling can inflate counts. The conversation stresses the rising cost of token usage for coding agents and advocates for techniques like dynamic context pruning to manage long-horizon tasks. Users share concerns about efficiency across platforms and seek practical token-saving methods for coding workflows.
Question for the room: What token-saving technique or tooling have you found most effective for coding agents in your projects?
Add a comment
0/2000