Token Insights

Cost / Total / Share — Repriced using real OpenRouter paid rates for each model (even "free" ones). Total = input + output tokens. Share bar = % of this period's total cost from models list.

Rate — $ per million total tokens (lower is cheaper per volume).

Limits — Typical public/free/dev tier (account-specific in reality). NVIDIA free on build.nvidia.com is base ~40 RPM (account-level; many users request increases to 100-200 RPM). No hard daily token limit documented for free tier, only RPM. OpenRouter free models are very low RPD.

Filter + Sort — Live filter by name. Click any header (Model/Calls/In/Out/Total/Cost/Rate) to sort; click again to reverse. Especially powerful on All Time.

Bars — Percentage of total spend for this period

Turns — Total conversation turns in this category

Edit Turns — Turns where file edits were made

One-Shot Rate — % of edit turns that succeeded on the first attempt, with no follow-up fixes needed. 100% one-shot means every edit was right the first try. No badge = no edits in that category (read-only).

Plan Mode — Models ranked for architecture/planning tasks (reasoning depth, context window, instruction following).

Build Mode — Models ranked for coding/execution tasks (speed, tool-use fidelity, code correctness).

Health — 🟢 Healthy, 🟡 Rate Limited, 🟠 Capped (daily limit), 🔴 Error/Unavailable, ⚪ Unknown.

Limit Type — Rate = recovers in minutes (RPM), Cap = daily limit with ~24h reset.

Usage/hr — Estimated usable requests per human hour (higher = more capacity).

Reset — Countdown to rate limit/cap reset (if applicable).