Google Unveils Gemini 4 Argon, Claims Benchmark Lead Over OpenAI and Anthropic
Google has unveiled Gemini 4 Argon, a frontier AI model that it says leads or ties rivals on 13 of 18 disclosed benchmarks. It is initially limited to trusted cyber defenders and select testers, with broader availability planned later.
VentureBeat reported that Google does not claim Argon wins every benchmark. In the benchmark table included in Google's embargoed materials, Argon records the highest score, or ties for the highest score, in more categories than OpenAI's GPT-6 Astra or Anthropic's Claude Opus 5.5. Across the 18 benchmarks Google disclosed, Argon leads outright on 12 and ties for first on one. GPT-6 Astra leads outright on three and ties Argon on one. Claude Opus 5.5 leads outright on two.
That gives Argon the largest number of benchmark leads or top scores in Google's comparison set, making it the leading frontier model by that measure. The results still show a close race. VentureBeat reported that OpenAI and Anthropic retain advantages in several important technical categories.
The claim is narrower than best model across all tasks. Argon appears to have the broadest top-score profile among the three frontier models in Google's disclosed comparison, but GPT-6 Astra remains ahead in several software, science-terminal and computer-use tasks, while Claude Opus 5.5 remains ahead in terminal-agent and post-training workflows.
For enterprise buyers, VentureBeat reported, model choice remains dependent on workload, even as Argon gives Google its strongest claim yet to overall frontier leadership by benchmark count.
Google said Argon also expands its output ceiling. The model supports 1 million output tokens, up from a previous limit of 64,000 tokens. The company described the capacity as industry-leading. VentureBeat reported that the longer output limit matters for agentic software engineering, audit, migration and legal-review workloads, where a model's value often depends on how long it can sustain a chain of work before returning control to a human.
The initial release to cyber defenders and pre-release testers means most users will not have immediate access. Google has said broader availability is planned later, without specifying a date in the disclosed materials.