Power user runs most tokens on open models via LunaRoute's concurrent-lane billing, not Opus
dbreunig · x · 2026-11-08
dbreunig shares his token usage breakdown: despite loving Opus 5.5, most of his tokens go to less demanding work — a "software factory" pipeline running ds4-flash and glm 5.3, plus a tweaked Hermes agent on meta spark.
These open models run on LunaRoute, an inference provider that bills by concurrent lanes rather than per-token, which he calls amazing. A textbook case of tiered model routing: frontier models for hard problems, cheap open models on flat-rate lanes for bulk tasks.
Related event: Blogger Mixes Open-Source Models to Cut Token Costs(2 posts)→
More from coding & agent
- 1,400 Claude Code sessions suggest Opus 5.5's reasoning shifted abruptly on Nov 3 — churchkey · 2026-11-08
- Claude's 'this will take months' estimates may be right, given the code bloat it produces — itigges22 · 2026-11-08
- Redditor builds a playable pixel world inside ChatGPT's page with a 37k-line userscript — ehtio · 2026-11-08
- "Your agent's favorite sandbox": a meme nails how to market Exo's hidden per-minute VMs — dbreunig · 2026-11-08
- Open-source MCP server lets Claude Code drive, fly and play GTA V in the background — Wonderful-Ad-8579 · 2026-11-08
- Harness leaderboard shows OpenCode lagging; author suggests Pi as faster alternative — mariofilhoml · 2026-11-08