Google's Argon beats Opus 5.5 on LiveBench reasoning, not coding

Which Google plans include Gemini 4 Argon?

Google says Google AI Pro and Ultra subscribers get Argon in the Gemini app and AI Mode in Search, that developers build with the Gemini API paid tier, and that Ultra subscribers can also try it in Google Antigravity and AI Studio. On October 15, Google's Antigravity post said Argon was rolling out there for Ultra subscribers; by October 16 it said Pro and Ultra. Google's API pricing page says 'Gemini 4 Argon is not available on the Free tier.'

How does Argon's access for cyber defenders differ?

Google says members of its Fairwind Program, for trusted cyber defenders, get Argon without cyber guardrails on a rolling basis, as do its own internal teams. Argon began rolling out to Fairwind members on September 30, and on October 13 Google began expanding the program to hundreds of additional partners. Google's pricing footnote applies to 'the Argon model intended for public access only.'

What does Google claim for Argon on other benchmarks?

Google says that, relying on pure reasoning, Argon scores 87% on the Embodied Reasoning (ERQA) benchmark, ahead of all other frontier models, and that it leads Gray Swan's Indirect Prompt Injection benchmark, cutting attack success across 15 adaptive attempts to 0.7%. It also says shortcut attempts fell to 2% on coding benchmarks where models have no incentive to hold back, its best performance compared to previous Gemini models.