Gemini 4 Hydrogen places fourth on AA-Robotics
Google's Gemini 4 Hydrogen is the top Google model on Artificial Analysis's AA-Robotics v0.1 benchmark but ranks fourth overall, behind Anthropic's Claude Fable 5.2 and two OpenAI GPT-6 models.
Google's newest model ranks behind Anthropic and OpenAI on Artificial Analysis's early-stage robotics benchmark. In a post on X, Artificial Analysis said Gemini 4 Hydrogen is the leading Google model on AA-Robotics v0.1 and fourth on the overall leaderboard.
The three models ahead of it are:
- Claude Fable 5.2 — 75
- GPT-6 Astra — 69
- GPT-6.1 Sol — 65
As we reported earlier, Fable 5.2 holds the top spot on the benchmark. This is the first AA-Robotics placement for Hydrogen, which Google launched recently.
The post does not give Hydrogen's score, so its distance behind GPT-6.1 Sol is unknown.
Artificial Analysis also made a qualitative observation. It said Hydrogen "was very honest about its performance": the model reported a minor spill while pouring a drink and admitted it had left a lid unclosed. The post doesn't say whether AA-Robotics scores that kind of self-reporting or whether other models flagged their own errors the same way.
The benchmark is still at version 0.1. The post doesn't describe its task set or scoring method, or say whether the rankings will change as the benchmark develops.