What the AI coding tools changed last night, read from the vendors’ own pages, price changes, withdrawn tiers, acquisitions and shutdowns, each with a link to the evidence, plus a board ranking the models developers actually run, computed nightly from real OpenRouter usage.
On its own page, read 14 September 2026
Which coding model Cursor, Claude Code and Copilot are actually running tonight
Which coding model the tracked coding tools are running tonight, ranked on adoption and re-counted every night.
- Scored on what developers actually run inside their coding tools, re-counted every night.
- 6.75T tokens through 6 tracked coding tools · $0.24 per million tokens blended · Under $1 / M band · 1311K context. Ranked on usage, which measures adoption rather than capability.
- 5.07T tokens through 6 tracked coding tools · $0.05 per million tokens blended · Under $1 / M band · 1311K context. Ranked on usage, which measures adoption rather than capability.

What the lab wrote about it, newest first
