Google announces Gemini 4 Argon but keeps it unavailable to the public
Google has announced Gemini 4 Argon, a new AI model positioned as a successor to the previously promised Gemini 3.5 Pro. The company says Argon delivers industry-leading results in coding, knowledge work, and cybersecurity, although the model is not currently available for public use. Google said its engineers are already using it extensively internally.
According to Google, Argon analyzed “fleet-wide telemetry data” and helped the company save 300 TiB of memory across its data centers. Its agents have also been used in efforts to migrate C/C++ codebases to Rust. Those projects include thousands of lines in the re2 and libgav1 libraries and more than 800,000 lines in the Fuchsia OS Zircon kernel.
Google cited several benchmarks in support of its performance claims. On the software-engineering DeepSWE v1.1 test, Argon scored 77.9 percent, ahead of GPT-6 Astra, Fable 5.1, and Opus 5.5. The company also described the model as a leader on the economic-analysis Vals Index test, which is intended to measure performance on long-horizon tasks.
Written by Sakhanda Wire AI from reporting by Ars Technica. Facts, quotes and full context are in the original article below.