Google announces Gemini 4 Argon frontier model
Google unveiled Gemini 4 Argon, a new frontier model rolled out first to a trusted group of cyber defenders through its Fairwind Program. The model is priced at $2 per million input tokens and $10 per million output tokens, with cached input tokens offered at a 95% discount. It supports up to 1M output tokens, a jump from the previous 64K limit, enabling deep, multi‑step reasoning.
Key points
- Gemini 4 Argon costs $2 per million input tokens and $10 per million output tokens, cached input at 95% off.
- Model supports up to 1M output tokens, up from 64K, and scores 77.9% on DeepSWE v1.1.
- Argon freed over 300 TiB of memory and tied for first on CWE‑bench v1 with a 68% score.
Early internal testing shows Argon delivering strong results across several benchmarks: 77.9% on DeepSWE v1.1, a leading 51.3% score on AutomationBench, and 91.7% on LVBench for long‑video understanding. In quantum algorithmic optimization it beat the published baseline by 40%, and its memory‑efficiency agents have already freed over 300 TiB of memory, with an estimated 500 TiB to 1 PiB in total savings. The model also tied for first on CWE‑bench v1 with a 68% score, demonstrating advanced vulnerability discovery.
Google is proceeding with a phased release, working with the U.S. government’s voluntary pre‑release process and strengthening safeguards against misuse, prompt injection, and misalignment. The company plans to expand access to developers, enterprises, and consumers after gathering feedback from the initial cohort of testers.
Model page: Gemini 4 Argon →
The story so far
2 episodes →- Google announces Gemini 4 Argon frontier modelthis story
Gemini 4 Argon: our next era of frontier intelligence
Google DeepMind · 30 September 2026
Loading the full article…
This text was published by Google DeepMind and written by Koray Kavukcuoglu. It is reproduced here with attribution so you can read it in full; the rights remain with the publisher. Read it at the source ↗
Coverage and discussion
2sources- Google announces Gemini 4 Argon AI model, but you can't use it yetPress · Ars Technica AI ·
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Generative AI & Models
All →- OpenAI releases GPT-6.1 Sol, nearly matching Astra at one-fifth the price · 24 src
- OpenAI scraps GPT-6.1 Astra release over safety concerns · 61 src
- InclusionAI releases Ling 3.1 Flash with 2-week free trial · 1 src
- Anthropic launches Claude Sonnet 5.5, claiming 30% faster output and up to 30% lower cost per task than Sonnet 5 · 25 src
- Anthropic launches Claude Opus 5.5 with 40% lower costs and Fable 5.1-level performance · 19 src
Comments
via GitHub Discussions