Noesion Radarnoesion.ai
Research2026-09-02

Google released ARC-AGI results for Gemini 3.8 Flash. It scored 98.5% on ARC-AGI-1 at high reasoning. It scored 89.2% on ARC-AGI-2 at high reasoning. It scored 10.4% on ARC-AGI-3 with Standard harness at high reasoning. It scored 35.0% on ARC-AGI-3 with Provider Adapter harness at high reasoning. Standard harness lets the model keep chosen notes across the environment. Provider Adapter harness preserves opaque reasoning state between requests. Both harnesses were tested across 25 environments. Results are verified by ARC Prize. The model card and blog post are linked from the results page.

What it unlocks

Higher scores on ARC-AGI benchmarks with reasoning variants and harness configurations.

  • 98.5% on ARC-AGI-1
  • 89.2% on ARC-AGI-2
  • 10.4% on ARC-AGI-3 Standard
  • 35.0% on ARC-AGI-3 Provider Adapter

Sources

Send this to someone who needs it

Shares the story and its sources. Nothing about you.

What does this mean for your job?

This is the story as everyone gets it. Once a week we send you the version written for your role — what changed, why it matters for the work you actually do, and one thing to try. Free while we tune it.