dottxt tool-call index: worst provider endpoint breaks 15x more tool calls than best
remilouf · x · 2026-11-18
dottxt's weekly "tool-call weather" report (Nov 6-15) shows the provider matters as much as the model: the worst endpoint can break 15x more tool calls than the best for the same model.
- New: Inkling 2 Small (ThinkyMachines), Nex-N2.6, ERNIE-5.5-Flash-Preview (Baidu)
- Worse: DeepSeek V4.1 Flash 1017 on AtlasCloud jumped from 1.7% to 17%; DeepSeek V4 Pro 0423 on Azure rose 3.3% → 11%; GLM 5.3 Flash on Novita hit 2.3%
- Better: gpt-oss-120b on CoreWeave dropped from 50% to 6.6%
The index tracks 64 models across 65 providers and ships an npx-installable skill so agents can auto-pick the best endpoint.
More from coding & agent
- The killer app of agentic programming is building harnesses for agentic programming — danbri · 2026-11-18
- Dev shows agentic video editing with HyperFrames + Claude, ditching FFmpeg pain — toolstelegraph · 2026-11-18
- ETH's AuXSpec: executable postconditions as RL reward, 27B model beats frontier LLMs at SpecGen — eth-algopt · 2026-11-18
- Running Gemma inside an office printer: 14MB RAM, 60s per token — TheTuringPost · 2026-11-18
- Dev runs Claude+Codex+Grok like a team, ships two releases in a day — GeneralDistinct5775 · 2026-11-18
- GAUSSIAN EVIL: a browser horror game homage to P.T. built with 3D Gaussian splatting — willeastcott · 2026-11-18