Systima · 2026-07-12 · notable
Systima — Claude Code sends 33k tokens before your prompt, OpenCode sends 7k
Systima ran Claude Code and OpenCode against identical tasks and measured the token spend. Claude Code injected 33k tokens of scaffolding on Sonnet 4.5 before the user prompt arrived; OpenCode used 7k for the same job.
Systima's teardown finds Claude Code eats a 4.7x token surcharge before the user prompt even arrives, and its cache breaks mid-session.
What is it?
The Systima writeup is a side-by-side measurement of Claude Code (Anthropic's coding CLI) and OpenCode (the open-source competitor) on identical prompts and infrastructure. Systima, an EU AI Act compliance consultancy, published it on July 12 and it hit 362 points on Hacker News the same day.
How does it work?
The team asked each harness for a one-line reply on Sonnet 4.5 and counted tokens before and after adding real production config (instruction files, MCP servers, subagents). Claude Code shipped roughly 33,000 tokens of system prompt, tool schemas, and scaffolding before the user prompt arrived; OpenCode shipped about 7,000 for the same work. Claude Code also rewrote its prompt prefix mid-session, so the KV-cache broke and the team saw up to 54x more cache-write tokens than OpenCode, billed at premium rates.
Why does it matter?
Coding agents charge by the token, so a 4.7x overhead on Sonnet 4.5 (3.3x on Fable 5) is a real bill difference, not a benchmark curiosity. Once the writeup adds production settings — instruction files and MCP servers push both harnesses up by 20k–25k tokens per request, and two subagents multiply consumption from 121k to 513k — the choice between Claude Code and OpenCode becomes a monthly-invoice question for any team running coding agents at scale.
Who is it for?
engineering leads picking between Claude Code and OpenCode for production coding-agent work
Try it
systima.ai/blog/claude-code-vs-opencode-token-overhead