Gemini 3.7 Flash scores 84.6% on ARC-AGI-2 at $0.25 per task
typewriters · x · 2026-08-21
ARC Prize reports verified results for Google's Gemini 3.7 Flash: 84.6% on ARC-AGI-2 at $0.25 per task, and 95.5% on ARC-AGI-1 at $0.12 per task. Relative to other frontier models, it stands out for combining high scores with very low cost.
Related event: Gemini 3.7 Flash Passes ARC-AGI Verification at $0.25 Per Task(3 posts)→
More from Models
- OpenAI only just added a pause; Google's approach differs, dev notes — BlackHC · 2026-08-21
- Science Benchmarks Crawl at 1-2% Monthly Progress; SciCode Saturation Is Far Off — Worldly_Beginning647 · 2026-08-21
- 'End of VC communism': investor slams OpenAI and Anthropic pricing talk — StewartalsopIII · 2026-08-21
- StepFun 3.5 was impressive; 3.7 disappointed, hoping for a mature return — teortaxesTex · 2026-08-21
- Open Weights Models Face UX Friction; Both Closed and Open Will Grow — JosephJacks_ · 2026-08-21
- OpenAI's 'math breakthrough' criticized: real work is prompt engineering — gerardsans · 2026-08-21