NVIDIA's AVO agent scores 100% on ARC-AGI-3, beats FlashAttention-4 in kernel optimization
MickeySteamboat · x · 2026-08-22
NVIDIA's new AVO agent architecture achieved 100% on the ARC-AGI-3 public set, completing all 183 levels across 25 environments. While the Claude Opus 5 model baseline scores around 30%, integrating it into AVO with persistent memory, supervision, and tools yields perfect performance. Previously, the same architecture ran autonomously for 7 days to optimize GPU kernels, exploring 500+ directions and producing kernels up to 10.5% faster than FlashAttention-4. This demonstrates that the AI frontier is shifting from the model itself to the entire agent system built around it.
Related event: NVIDIA's AVO Agent Scores 100% on ARC-AGI-3 Public Set(18 posts)→
More from coding & agent
- Lovable gets slow as project scales; seeking faster alternatives — djays1618 · 2026-08-22
- Ox Alpha Generates 64k Token 3D World in One Shot — rohanpaul_ai · 2026-08-22
- Terminal-bench Results: GLM 5.3 and Fable 5 Show Varied Performance — mariofilhoml · 2026-08-22
- Two Ludi agents talking to each other after a tough demo day — Kangwook_Lee · 2026-08-22
- Qwen Code v0.21.14 nightly released: fixes CI, Web Shell OOM, and adds runner isolation — qwen-code-ci-bot · 2026-08-22
- Dev uses Codex to automatically scout cheap GPUs and build a dashboard — majidmanzarpour · 2026-08-22