Testing 105 Bugs: Luna Beats Fable in Cost-Efficiency, Expensive Models Avoid Coding
PawelHuryn · x · 2026-08-01
Developer PawelHuryn shared an AI model routing strategy based on 105 hidden bugs, 9 frontier models, and 14 test runs:
- Judgment, strategy & complex cases: Fable 5
- Frontend: Opus 5, followed by Kimi K3
- Writing: Opus 5
- Coding & debugging: Luna (max effort), Sol (high effort), or Grok 4.5
Cost-efficiency: Luna at max effort fixed 33 bugs for $1.80, while Fable 5 fixed 29 for $104.49. The author notes that the most expensive model he uses never touches code.
Related event: Dev Open-Sources Bug-Hunt-Bench to Test LLM Debugging(2 posts)→
More from coding & agent
- Tencent releases Hy-Memory-30B-A3B, a 3B-active memory-focused model claiming to beat Opus — victormustar · 2026-10-16
- Founder wipes 30k lines of tech debt in one shot with Opus — lucasmeijer · 2026-10-16
- Japanese engineer runs an AI agent org inside tmux: per-project teams, one human talking to the orchestrator — 4310sy · 2026-10-16
- gloria-mcp ships an MCP server serving Gloria AI's curated crypto news — modelcontextprotocol · 2026-10-16
- Pine Script MCP server lets AI agents read v6 docs and validate syntax — modelcontextprotocol · 2026-10-16
- Open-source ebook2audiobook turns any ebook into a free audiobook, 20.4k GitHub stars — thisguyknowsai · 2026-10-16