Google announces Gemini 4 Argon as its new frontier model

Google has unveiled Gemini 4 Argon, its newest frontier model, heralding a new era of deep reasoning and complex problem-solving. This comes after the cancellation of Gemini 3.5 Pro and a shift towards 3.8 Flash. The new model, featuring a unique naming scheme, is poised to deliver frontier-level capabilities in coding, knowledge work, cybersecurity defense, and creative writing.

One of its standout features is a 1M token output limit, a significant increase from its previous 64K limit, enabling the processing of longer and more intricate tasks.

In addition to its expanded reasoning and coding capabilities, Gemini 4 Argon boasts enhanced performance in cybersecurity defense, with a DeepSWE v1.1 score of 77.9%, surpassing competitors like Claude Opus 5.5 and GPT-6 Astra. Initially set to be available without cyber guardrails to trusted defenders and internal Google teams, the model is expected to soon roll out to Google AI Ultra subscribers and paid API customers.

Google's internal teams have already begun leveraging Gemini 4 Argon for various tasks, including memory optimizations across Google's data centers, freeing up over 300 TiB of memory. The model is also being used for migrating large-scale codebases to Rust, enhancing the performance of systems like Google's video decoder, which now runs 2.7 times faster than its previous version while maintaining identical video output.