For anyone getting hit by downgrades or random bans: PGS has 0731, 4.1 and GLM 5.3 flash free for a month (with memory, voice, skills etc)

We've been hearing a lot about people getting locked out, or experiencing quality dumps across lots of platforms. It seems to be happening across the board with usage tightening and models downgrading silently. Along with all the drama surrounding open source AI in the USA, it's more important than ever that we all have real access to the models, reliably. Western closed AI acts like it has a monopoly on quality app features: memory, skills, voice, canvas etc. Meanwhile memory stays locked in. Models get swapped or "updated" into downgrades. Privacy changes from service to service and ads are starting to show up. The whole consumer side experience is extractive. We're trying to prove that you can actually have great, reliable model access, privacy, and all the bell and whistles of big AI apps. all the best open models in one place, on private US infrastructure, with the complete app experience around them. Fully private, direct service. It should be, and can be, that simple. What that means in practice: The models we offer: DeepSeek, GLM, Kimi, Minimax, Nemotron Ultra and more, side by side in one app. Hop models mid conversation if you feel like it. No digging through five different apps and API dashboards to reach the models you actually like. Actually private: US based processing and your conversations are never used for training. Ever. That is the entire point. These labs open sourced incredible models and we think you should get to use them without handing over your data as the price of admission. Real memory that learns and grows with you: Not a context window that fills up and dumps you. Persistent memory that carries across conversations, fades gracefully when unused, and wakes back up when it's relevant again. There's even a nightly consolidation pass, the system basically sleeps on it and writes down what mattered. Quality Voice Mode: Yes, actual voice mode with more than a dozen voices on open models. Bring your history. Coming over from ChatGPT, Claude, or Gemini? Export your chats and import the whole thing, it becomes live memory on day one. You can literally just zap your chat history from a backup file and have all your chats waiting for you. Multiple nodes. Separate workspaces with separate memories, so your coding setup doesn't share a brain with your journal. You can set a node for work, a node for personal, or whatever. Each one remembers your choices for voice, font, memory etc. so you can have totally different setups and agents to fast switch between. Genuine thanks to DeepSeek labs. They've been doin an amazing job releasing some of the best models on the planet! We'll always appreciate the work it takes to create these amazing models! Open Grove is here and It's free for a month if anyone want's to check it out (or just use the models for free for a while): pgsgrove.com/open-grove-overview We're finding the DSV4 updates and GLM 5.3 models with memory to be mindblowing. Recognition and long context reasoning are fantastic, and the memory evolution with the particular models is amazing.