Google Cloud switches AlphaEvolve default to Gemini 4 Argon and adds OpenAI GPT-6 models
Google announced that starting October 26, any AlphaEvolve experiment on Google Cloud that leaves the model field “Unspecified” will automatically run on Gemini 4 Argon, replacing the previous Gemini 3.8 Flash default. The change only applies to experiments created after that date; earlier experiments keep their recorded model. Because Argon is a higher‑tier model, token pricing will also shift,…
Key points
- From Oct 26, AlphaEvolve experiments without a specified model will default to Gemini 4 Argon, replacing Gemini 3.8 Flash.
- In Japan, Gemini 2.5 Pro is discontinued for domestic Gemini Enterprise apps and switched to Gemini 3.5 Flash.
- OpenAI’s GPT‑6 Astra, GPT‑6.1 Sol and GPT‑6 Luna are now available on Google Cloud via invite‑only preview, with GA expected in weeks.
For users in Japan (and Canada), Gemini 2.5 Pro has been discontinued for Gemini Enterprise apps that store data locally, and those workloads have been switched to the domestic Gemini 3.5 Flash version. The model remains available in global, US and EU regions.
OpenAI also revealed a partnership with Google Cloud, making GPT‑6 Astra, GPT‑6.1 Sol and GPT‑6 Luna available in the Model Garden on an invite‑only preview. General availability is expected within the next few weeks, and usage will be billed through the Google Cloud account, simplifying contracts for customers who primarily use Google Cloud.
Model pages: Gemini 4 Argon → · GPT-6 Astra → · GPT-6.1 Sol →
The story so far
3 episodes →- Google Cloud switches AlphaEvolve default to Gemini 4 Argon and adds OpenAI GPT-6 modelsthis story
Today's Cloud AI Memo (10/25): 'Default Models' are changing starting tomorrow—AlphaEvolve's switch to Gemini 4 Argon and OpenAI models coming to Google Cloud
note.com · 24 October 2026
Loading the full article…
This text was published by note.com and written by ルルー. 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 →- Anthropic investigates elevated errors across five Claude models · 1 src
- Moonshot AI releases Kimi K3.1 Pro, a 2.8T open-weight model closing gap with proprietary leaders · 7 src
- Llama.cpp PR adds DeepSeek-V4.1-Flash support with review fixes · 1 src
- Claude Opus 5.5 users report cutoffs on reverse-engineering prompts · 1 src
- Block releases Dessa, a full-duplex speech model trained on human conversation · 4 src
Comments
via GitHub Discussions