antirez.bsky.social

In the glm5.2 branch of DwarfStar you can find a preview of GLM5.2 support. The model is strong and works well, but I'm highly hesitant to say that the 4bit GLM5.2 quants are able to perform *strongly* better than DeepSeek v4 Flash, which has a decisive speed advantage: github…

GitHub - antirez/ds4 at glm5.2 github.com

What Europe should do right now: 1. Call all the European researchers working on AI and return them back with same salary (or they can stay but switch career). 2. Fill EU places having GPUs with money, and put those people there. 3. AI partnerships with China + India.

Another important thing: Chinese models are not strong because they distill US models. Distillation of models via API is *impossible*. If somebody tells you the contrary, they don't understand machine learning:

Modern AI resulted from research made also by many non-US scientists (Hinton, the French folks, Linnainmaa, many others). The pre-training corpus was produced worldwide with massive code contribution from Europe OSS. What is happening with frontier LLMs is unacceptable.

Big news for DwarfStar users: I got DeepSeek v4 Flash and GLM 5.2 working with Tensor Parallelism across 2 M5Max 128GB MacBooks via RDMA. It is especially interesting for GLM since otherwise, fully resident, can't fit a machine that money today can buy... Now it can.

Today I had an harder than usual question for my local model (security). With SSD streaming now DwarfStar can run DeepSeek v4 PRO at 4.15 t/s, and this was more than enough to get a detailed reply. I already feel "safer" than before in my AI future. M5 max 128GB, model 433GB.

DawrfStar with DeepSeek 4 Flash 4 bit, PRO 2 bit, and GLM 5.2 4 bit. M3 Ultra 512GB. Results on hard programming tasks with GPT 5.5 as a judge. So the GLM 5.2 branch is going to be merged and supported for CUDA + Strix Halo as well.

DeepSeek v4 will be updated mid-July. This is a very good news. Now: I hope GLM 5.3 will have the same architecture as GLM 5.2. Models are recently changing a lot because of big (to the attention) and small tweaks. This is good but costly development-wise.

DwarfStar now supports SSD streaming in the DGX Spark and Strix Halo, not just in Metal. You can run the Q4 quants at decent speed, and even DeepSeek v4 PRO at low speed, or you can run Q2 Flash if you have less than 128GB.

GLM 5.2 for some reason, while not a QAT, seems to be very resistant to extreme quantizations similarly to DeepSeek v4. Unexpected.

There is something I don't understand about Fable: they want to go b2b with Fable available only via API at that absurd price point. I believe this is delusional (OpenAI exists) but let's buy it for a moment. But: they ask for 50$ 1M tokens with this switch-back-to-opus trigger?