Google has announced Gemini 4 Argon, describing it as its "next era of frontier intelligence" and a model built for deep reasoning across complex, long-horizon workflows. The launch was reported by 9to5Google and independently confirmed by Android Authority, giving the new model a second source of confirmation.
The announcement introduces a new naming scheme and follows the cancellation of Gemini 3.5 Pro, with Google shifting its focus to 3.8 Flash in the interim. According to 9to5Google, Google is positioning Gemini 4 Argon for frontier-level capabilities across coding, knowledge work, cybersecurity defense and creative writing.
Gemini 4 Argon specs and capabilities
The headline technical change is the output token limit: Gemini 4 Argon can generate up to 1M tokens, up from 64K. Google says that headroom allows the model to think deeply and produce hundreds of thousands of tokens in a single trajectory, adding a new level of depth when solving tough problems in one go.
Google also points to expanded coding, reasoning and multimodal capabilities. In the benchmarks cited by 9to5Google, Gemini 4 Argon scores 77.9% on DeepSWE v1.1, ahead of Claude Opus 5.5 at 74.2% and GPT-6 Astra at 74.1%.
- Output token limit: 1M tokens, up from 64K
- DeepSWE v1.1 score: 77.9%
- Claude Opus 5.5: 74.2% on the same benchmark
- GPT-6 Astra: 74.1% on the same benchmark
