Google launches Gemini 4 Argon, skipping Gemini 3.5 Pro

By Splitfeed EditorialLast updated Oct 1, 2026, 11:00 UTC

On September 30, 2026, Google announced Gemini 4 Argon, its latest flagship AI model, skipping the Gemini 3.5 Pro release that many had anticipated. The model is priced at $2 per million input tokens and $10 per million output tokens, and it can generate up to one million tokens in a single response. Alphabet shares rose in premarket trading following the announcement. The launch places Gemini 4 Argon in direct competition with OpenAI's GPT-6 Astra and Anthropic's Claude Fable 5.1, with autonomous cybersecurity capabilities highlighted as a key differentiator.

The version skip from Gemini 2 through to Gemini 4 without a Gemini 3.5 Pro intermediate release is notable because it signals a deliberate rebranding or internal restructuring of Google's model roadmap, rather than a straightforward iterative update. The one-million-token output ceiling is a significant technical threshold, as it enables tasks like generating and reviewing very large codebases or documents in a single pass. The autonomous cybersecurity focus is also worth attention, since it moves AI capability into a domain with direct and sensitive real-world consequences, raising questions about oversight, liability, and potential misuse that go beyond typical model benchmarking debates.

None of the three camps, Pro-AI, Anti-AI, or Middle Ground, have published reactions to this story yet. Pro-AI voices would typically celebrate the technical leap, pointing to the one-million-token output and the competitive pricing as evidence that AI is becoming more powerful and accessible. Anti-AI voices would likely raise concerns about the autonomous cybersecurity angle in particular, arguing that giving AI models greater autonomy in a high-stakes domain like cybersecurity introduces risks that outpace current regulatory frameworks. Middle Ground commentators would be expected to acknowledge the capability gains while pressing for clearer guidelines on how autonomous cybersecurity features would be governed and audited.

The details most worth watching in the near term include benchmark comparisons against GPT-6 Astra and Claude Fable 5.1 as independent evaluations emerge, any regulatory or policy responses to the autonomous cybersecurity capabilities, and whether Google provides further explanation for the version numbering jump, which may hint at discontinued internal projects or a broader product strategy shift.