# First impressions of Gemini 4 Argon

> I had Gemini 4 Argon review my AI agents, skills and blog drafts. It’s fairly slow, but strong on text; its explanations still need to be more direct.

Published: 2026-10-27 · Facts checked on 2026-10-27
URL: https://sobra.be/en/nostrabramus/gemini-4-argon/
Found via: https://tweakers.net/nieuws/253284/google-brengt-gemini-4-argon-na-uitstel-om-cyberzorgen-alsnog-uit.html

Google released [Gemini 4 Argon](https://blog.google/innovation-and-ai/models-and-research/gemini-models/introducing-gemini-4-argon/) widely on 13 October. It's available through the application programming interface (API), and Pro and Ultra subscribers can also use it in the Gemini app and AI Mode in Google Search.

I still need to test Argon on a real codebase, but I did have it review my own AI agents and skills for gaps and inconsistencies. That turned up genuinely useful feedback. I also ran it with my copywriting skills to review and rewrite these blog drafts. For writing and editing text, I find Argon better than Claude and ChatGPT, though it is noticeably slower than Claude and especially ChatGPT.

What also stood out to me is that Argon's replies are long and not immediately concrete. It asks before making changes, but I can't tell from a single line what it actually wants to change. I end up having to read everything carefully to be sure it's right, whereas one clear sentence is often enough to see whether a model has understood the point. I want to stay in control, not micromanage.

For actual coding, my gut feeling is still that Claude and Codex have the edge. Google's own [model card](https://deepmind.google/models/model-cards/gemini-4-argon/) shows a mixed picture too: Argon wins on some coding tests, Claude and GPT-6 Astra on others. Next up is testing how that holds up in an existing project.

## Sources

- [Gemini 4 Argon: built to solve your hardest problems (Google)](https://blog.google/innovation-and-ai/models-and-research/gemini-models/introducing-gemini-4-argon/)
- [Gemini 4 Argon model card (Google DeepMind)](https://deepmind.google/models/model-cards/gemini-4-argon/)