dottxt tool-call index: ERNIE-5.5-Preview breaks 0.74–1.20% of tool calls on Baidu
remilouf · x · 2026-11-20
- .TXT launched a Tool-call Index tracking broken tool-call rates across models and providers: ERNIE-5.5-Flash-Preview's launch week (Nov 10–17) shows 0.74% broken calls on Baidu (free) and 1.20% on Baidu (full precision).
- The index covers dozens of models including DeepSeek, Kimi, MiniMax, GLM, Qwen, Gemma, Llama, and Mistral, listing the best rate per provider to help pick stable tool-calling setups.
More from Models
- Coding Agent Index: Google tops leaderboard at 59 as Anthropic preps IPO — zacharynado · 2026-11-20
- Charted: Claude Fable 5.2 vs GPT-6.2 Astra vs Gemini 4 Neon intelligence-speed frontier — thecoolbrian · 2026-11-20
- Gemini API previews tool retrieval: schemas load on demand, cutting token cost — petrusenko_max · 2026-11-20
- "Neon gave me AI psychosis": researcher describes overconfidence spike from new top model — andrew_n_carr · 2026-11-20
- "Beating Gemini is the benchmark now": CursorBench chart mocked as benchmaxxing — zacharynado · 2026-11-20
- User Rants: Claude Now Returns Paragraphs That Answer Nothing — AccidentalFolklore · 2026-11-20