75M model hits 44% on ARC-AGI for just $0.67 training cost
Saboo_Shubham_ · x · 2026-09-02
GitHub project mdlARC demonstrates a highly efficient training approach: a 75M parameter standard Transformer trained on a single RTX 5090 for about 1.5 hours costs only $0.67, yet achieves a score of 44% on the ARC-AGI-1 public evaluation. The author claims this is one of the best non-LLM scores to date at an extremely low cost. The code is fully open-source.
Related event: 75M Model Trained for $0.67 Scores 44% on ARC-AGI-1(3 posts)→
More from coding & agent
- Anthropic borrows lawyers' ethics for Claude agents that closed 186 deals with no human approval — rohanpaul_ai · 2026-10-18
- One prompt turns Claude Code into a full motion-graphics video pipeline with Remotion — aziz4ai · 2026-10-18
- Prompt injection via email can hijack your AI agent — jev-guard adds risk-scored guards — sujingshen · 2026-10-18
- 14-Agent AI Film Pipeline on One RTX 5090: Fixing Cut Defects Between Clips — developervkmp · 2026-10-18
- Meta paper introduces "agent plasticity": best-performing agent isn't the most efficient learner — rohanpaul_ai · 2026-10-18
- Dev ships a bushfire prediction platform in a weekend with frontier models — CtrlAltDwayne · 2026-10-18