Google says Gemini 4 Argon saved over 300 TiB of data‑center memory
Google announced Gemini 4 Argon on September 30, 2026 and made it available to Google AI Pro, Ultra users, developers and businesses on October 13. The company says the model’s agents analyzed data‑center operations and freed more than 300 TiB of memory, with a longer‑term goal of cutting 500 TiB to 1 PiB. Google also reports that an AI‑driven rewrite of about 32,000 lines of the libgav1 video…
Key points
- Google says Gemini 4 Argon reduced data‑center memory usage by over 300 TiB.
- Gemini 4 Argon rewrote ~32,000 lines of libgav1 code, achieving a 2.7× faster video decoder.
- The model can generate outputs up to about 1 million tokens via a dedicated API setting.
According to Google, Gemini 4 Argon can produce outputs up to roughly 1 million tokens when specific API settings are used, far beyond the usual 64,000‑token limit. The model tops the “Vals Index” for real‑world tasks such as finance, law and programming, though the article notes that ChatGPT and Claude still outperform it on other benchmarks. The author, Naho Uemura, reflects on how such long‑context capabilities could help marketers and solo entrepreneurs automate research, content revision and training‑material creation, while cautioning that the claims are based on Google’s own data and have not been independently verified.
The story so far
3 episodes →- Google says Gemini 4 Argon saved over 300 TiB of data‑center memorythis story
300 TiB of memory saved with Gemini 4 Argon: AI evolution seen through Google's real-world examples
note.com · 18 October 2026
Loading the full article…
This text was published by note.com and written by 上村菜穂 株式会社PR NET・CEO. It is reproduced here with attribution so you can read it in full; the rights remain with the publisher. Read it at the source ↗
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 makes Prompt-to-Voice generally available for developers · 3 src
- Meta releases Muse Spark Open, a 816B-parameter multimodal model with 1M-token context · 37 src
- DeepSeek releases DeepSeek-V4.1-Flash-1017, scores 77.0 on DeepSWE v1.1 and opens Search and Fetch APIs · 19 src
- Anthropic’s Claude Opus 5.5 tops Arena AI model rankings · 2 src
- Modders report DLSS 5 yields up to 90% performance boost on unsupported games · 2 src
Comments
via GitHub Discussions