Hey guys, I'm currently considering upgrading to Google ai Ultra x20 because I found the new Gemini 3.7 Flash model quite decent and fast. What was your experience with the Google AI Ultra plan ? Are the usage allowances on the Gemini models generous? Are they also generous for Claude Opus 4.6? Are there any other special features worth noting?
When can we expect Minimax M3 Pro 2.7T to drop? Minimax announced that it was expected to be released in Q3 2026 and that it's still in development.
I'm currently using the new Dsv4 pro GA on max reasoning through the commandcode API. The issue im facing is, that after it reasons for a while (1000+ lines which it does often) it randomly stops and im getting the error below. Then the model starts reasoning again, completely from the start and all the tokens before are wasted. This happened in the Commandcode CLI too everytime i tried it. In the commandcode CLI it is showing as a disconnect. My internet connection is perfectly stable so that shoouldnt be the reason. I'm currently using the Commandcode plan. I hope that this issue can be resolved since the Ds v4 Pro GA model is often unusable this way since it keeps thinking again after it gets interrupted from the error.
What's your experience with Deepseek VR Pro GA so far? What have you been building with it?
Are they worth purchasing? Is 800 requests per 5 hours enough to run one Opencode agent?
I'm thinking about switching to a Kimi subscription now that they have a frontier model.
I'm new to all of this so I'm trying to find a good "agent harness" that works well. I've already tried using the go models in VS code Copilot through an extension and that worked pretty well for me. I'm just curious if there are any other free options that work even better. Thanks for the help In advance!
Dm me if you got some.
