AutoRouter launches beta: auto-routes each coding task across Claude, GPT and other models to cut token spend
floguo · x · 2026-10-23
AutoRouter has opened its beta as a "decision layer" for AI coding tools: it reads each prompt, picks the best-fit model, and plugs into your existing setup via the model menu in Codex or /model in Claude Code.
How it works
- Works with your existing ChatGPT and Claude plans plus pay-per-use models (Gemini, Grok, DeepSeek, Kimi, MiniMax), routing hard tasks to frontier models on your plans and trivial work (commit messages) to free ones
- Every model must clear a quality bar before price is considered; unclear tasks go to a strong model, never the cheapest
- Supports custom rules (e.g. always route UI work to Claude Opus 5.5) and manual overrides; new models appear in the picker immediately
- Every turn logs the model, cost, and routing rationale; a dashboard shows savings vs. top model
Official example: a 9-minute, 126.5k-token session stayed entirely within plan quotas, saving $2.41 versus always using the top model. A cross-file refactor routes to Claude Opus 5.5, a flaky CI-only test to GPT-6 Astra.
More from coding & agent
- NVIDIA Launches Verified Agent Skills to Bring Capability Governance to the Skill Layer — bradneuberg · 2026-10-23
- Cognition open-sources Devin agent plugin to patch the day's CVE wave — Scobleizer · 2026-10-23
- Fine-tuning models solo with a Grok bot and rented GPUs — Rasmic · 2026-10-23
- One prompt found pre-auth PHP RCEs: agent-guided vuln hunting timeline — chrisrohlf · 2026-10-23
- How Speak uses GPT-Live and a Decisions API for ultra-responsive live AI language tutoring — pbbakkum · 2026-10-23
- Google Live API adds decisions model triggers, faster than function calls for voice agents — pbbakkum · 2026-10-23