Optimizing Claude subagents to reduce high token costs
Nearby_Life9503 · reddit · 2026-08-31
A user experimenting with Claude CLI subagents noticed that creating sub-processes incurs high token costs because the system sends context, cache, and other info with every sub-agent creation. The current setup uses Opus 5 high for the main model and Opus 5 low for sub-agents. The user is seeking optimization strategies or advice on whether switching to a single model (like Opus 5 medium) would be more cost-effective.
More from coding & agent
- Open-Source GitHub AI Code Reviewer: Self-Hosted Models, No SaaS Fees — Arindam_1729 · 2026-08-31
- Library recreates Claude Code, Codex, and Grok interfaces as shadcn components — RichardsonDx · 2026-08-31
- The Post AI Data Stack: Infrastructure Challenges — sh_reya · 2026-08-31
- OpenClaw 2.0 Released: Rebuilds Installation and Browser Experience — heyneighbor · 2026-08-31
- MCP Server Lets LLM Agents Play Civilization VI Natively — jia_seed · 2026-08-31
- Hermes Agent running GLM-5.2 via OpenRouter on a 2GB VPS — granawkins · 2026-08-31