Google's Gemini 4 Argon Cracks ARC-AGI-3 but Trails in Math and Coding

1 reports archived1 independent sourcesupdated 10/16/2026, 23:03:46

Synthesis & Latest Updates

1 Sources Cross-Validated

Google DeepMind's new frontier model, Gemini 4 Argon, scored 97.5% on the ARC-AGI-3 benchmark, becoming the second lab to cross the 95% threshold after OpenAI. However, Epoch AI notes that this result heavily depends on a specific testing harness using vendor features, while the model ranks only eighth in math and sixth in software engineering, revealing significant capability gaps.

LATEST/Google DeepMind's new frontier model, Gemini 4 Argon, scored 97.5% on the ARC-AGI-3 benchmark, becoming the second lab to cross the 95% threshold after OpenAI. However, Epoch AI notes that this result heavily depends on a specific testing harness using vendor features, while the model ranks only eighth in math and sixth in software engineering, revealing significant capability gaps.

TIMELINECoverage timeline

Total 1 reports · Latest first

The DecoderT2·Oct 16, 23:0378 pts
Google's Gemini 4 Argon Cracks ARC-AGI-3 but Trails in Math and Coding
- Gemini 4 Argon scored 97.5% on ARC-AGI-3, trailing only GPT-6 Astra at 99.9%.
- The high score relies on the Provider Adapter Harness, which utilizes vendor-specific features like retained reasoning and context compaction.
- While leading in knowledge, agent, and vision benchmarks, it ranks 8th in Math ECI and 6th in Software Engineering Index.
View Related Reports & Analysis Original Source