Claude Code sends 33k tokens before reading the prompt; OpenCode sends 7k

Claude Code Sends 4.7x More Tokens Than OpenCode Before Reading Your Prompt

systima.ai

July 12, 2026

17 min read

🔥🔥🔥🔥🔥

71/100

Summary

Claude Code uses approximately 33,000 tokens of system prompt, tool schemas, and injected scaffolding before processing a prompt, compared to OpenCode's 7,000 tokens. This results in Claude Code sending 4.7 times more tokens than OpenCode before reading user input.

Key Takeaways

Claude Code uses approximately 33,000 tokens of system prompt and related overhead before processing a prompt, compared to OpenCode's 7,000 tokens.
Claude Code rewrites tens of thousands of prompt-cache tokens mid-session, resulting in up to 54 times more cache token usage than OpenCode.
A production repository's instruction file can add an average of 20,000 tokens to every request, leading to a total of 75,000 to 85,000 tokens sent before user input.
For multi-step tasks, Claude Code can result in lower total token usage than OpenCode by batching tool calls into fewer requests.

Read original article

View story →

Coverage & discussions

Hacker News · systima.ai

Claude Code sends 33k tokens before reading the prompt; OpenCode sends 7k ↗

Community discussions

Hacker News discussion ↗

Historical count: 5 · 5 saved in sample · Historical sample

Analysis ready

Community Sentiment

Negative

Positives

Fable's 'curiosity' could actually enhance complex project planning, allowing subagents to explore multiple angles for thoroughness — a game changer for ambitious tasks.
Using a minimal system prompt with Claude Code can significantly streamline interactions, tackling inefficiencies head-on and potentially saving users from excessive token consumption.

Concerns

Claude Code's token consumption is outrageous, with subagents burning through budgets before even starting real work — it feels like a ploy to maximize profits at users' expense.
There's a clear negative incentive for Anthropic to keep users reliant on expensive subscriptions rather than addressing inefficiencies, making it seem like they're more interested in profits than user experience.
Token inflation is a real issue; simple tasks now require far more tokens than before, leaving users frustrated with the escalating costs for basic functionality.