Why tool calls break differently across models, and the simple fix a team used
Alternative-Swing-63 · reddit · 2026-07-25
A team found that tool calls failing across models is not random: each provider breaks in its own predictable way.
- On OpenAI reasoning models, certain disliked schema properties cause hard errors.
- Gemini may silently ignore constraints instead of failing loudly.
- DeepSeek and Llama may simply skip tool calls.
- Their fix was practical: move constraint text into the property description rather than fighting prompts.
Related event: Testing 16 LLMs Reveals Schema Issues Behind Tool Call Failures(2 posts)→
More from coding & agent
- theo publishes tstack skills, Garry Tan urges everyone to 'publish your stack' — ycombinator · 2026-10-26
- Agent verified 'success' after double-voting: a lesson in harness vs model bugs — TejasKumar_ · 2026-10-26
- diginomica: agentic automation hype is overdone, and 'The Hackening' of enterprise security has begun — jonerp · 2026-10-26
- 'Vibe verification': ~60% of AI-written formal specs prove the wrong property — basedjensen · 2026-10-26
- Devin launches Security suite powered by Claude Mythos with $100M+ credits — DevinAI · 2026-10-26
- HackBench injects reward hacks into agent trajectories to stress-test monitors — DanielKhashabi · 2026-10-26