labs/aicharts: fast model for specs #39

Merged
teahubxyz merged 1 commits from labs/aicharts-model into main 2026-10-13 14:01:34 +00:00
Owner

Live: simple prompts were fine, harder ones took 70s+ and came back empty (gemma-4 spends its budget thinking). Default the demo to @cf/meta/llama-3.3-70b-instruct-fp8-fast (AICHARTS_AI_MODEL still overrides), max_tokens 900.

🤖 Generated with Claude Code

https://claude.ai/code/session_01VDkN7ukK1kXkqov8FaZxva


Generated by Claude Code

Live: simple prompts were fine, harder ones took 70s+ and came back empty (gemma-4 spends its budget thinking). Default the demo to `@cf/meta/llama-3.3-70b-instruct-fp8-fast` (AICHARTS_AI_MODEL still overrides), max_tokens 900. 🤖 Generated with [Claude Code](https://claude.com/claude-code) https://claude.ai/code/session_01VDkN7ukK1kXkqov8FaZxva --- _Generated by [Claude Code](https://claude.ai/code/session_01VDkN7ukK1kXkqov8FaZxva)_
Sign in to join this conversation.