If you have ollama running be sure to try Gemma 4 31B Just released by Google. Much better than expected and runs much lighter than expected. Running in full thinking mode at less than 20G usage on my mini. Performs high in tests. Tool calling, image reading, ocr, full thinking. Solid model for replacing most of the simple stuff opus or sonnet were likely handling on the Claude plan. Stuff like checking emails and calendars and stuff would be set up easily I bet. Worth a look for sure. Quick bet is to build/install the OpenClaude harness while you have Opus to do it!

Replies (2)

Just realized I quoted Leon at the usage of <20Gb ram usage on my brief initial test. Tell me know if you measure otherwise. Leon also told me yesterday that my context window isage was 1,230,000 of 1,000,000 and I should expect compaction soon. 🤨 View quoted note →