Dev finds typesafeai too small for spec-quality assessment; Gemini Flash wins but costs

julianharris · x · 2026-09-19

julianharris spent two days deep in @typesafeai for content-quality assessment of software specs and found it miles short — clearly a small model lacking the needed intelligence. Gemini Flash 3.8 blows it away, but is too expensive for his taste. He's asking for alternative suggestions.

Related event: Small Models Fail Software Spec Evaluation as DeepSeek Flash Beats Gemini Flash(2 posts)→

Original post →

More from Models

Models channel →