AivexaNewsSearch
AI news for builders and product teamsChecked every hour

Gemini 4 Argon: our next era of frontier intelligence

Collected Sep 30, 2026

Google DeepMind announced Gemini 4 Argon, a new frontier model now rolling out to a set of trusted cyber defenders through its Fairwind Program. Koray Kavukcuoglu, SVP of Google DeepMind and Chief AI Architect at Google, described the model as built to sustain deep reasoning across complex, long-horizon workflows. Google said it is engaged in the U.S. government's voluntary process for pre-release model access and will expand access gradually while iterating on guardrails before making Argon available to developers, enterprises, and consumers.

Argon will launch at an introductory price of $2 per million input tokens and $10 per million output tokens, with cached input tokens priced at 95% off the input token price. Google said the model's output token limit is expanding to an industry-leading 1 million tokens, up from 64,000 tokens.

Google reported that Argon already powers internal workflows, citing quantum algorithmic optimization that beat a published baseline by 40%, agent-driven memory optimizations freeing over 300 TiB of memory with estimated total savings of 500 TiB to 1 PiB, and C/C++ to Rust migrations scaling up to more than 800,000 lines for the Fuchsia Zircon kernel. For libgav1, Google said Argon agents replaced 32,000 lines of SIMD code, producing a memory-safe video decoder that runs 2.7x faster than the existing Rust port with identical video output.

On benchmarks, Google said Argon sets a new state of the art on DeepSWE v1.1 at 77.9%, leads the Vals Index, ranks first on AutomationBench with 51.3%, and scores 91.7% on LVBench. In cybersecurity, Google said Argon is being released to trusted defenders and internal teams without cyber guardrails, ties for first on CWE-bench v1 at 68%, and that Wiz is already using it through its Scan for Good initiative, where it uncovered a critical vulnerability in healthcare software that previous frontier models had missed.

Google also detailed safeguards covering misuse prevention for cyber and CBRN attacks, prompt injection resilience, misalignment monitoring of chain-of-thought and actions, and hardened sandboxed environments. It said these safeguards underwent internal and external red team testing.

Read at Google DeepMind

Based on reporting from the original publisher. Visit the source for full context and later updates.

Publisher excerpt