What Ai model and hosting do you use?
I have a local Fedora workstation, it's nothing special and CPU bound, but it does have 32GB of RAM so allows me to run ollama with moderate sized local models. They run well, but slowly, my favourite right now is nemotron-3-nano 32B, but my Claw seems to prefer mistral 7B.
Most of the time TED uses the OpenAI Codex on the rate limited £20 a month plan, but he can now fall back to Minimax on ollama or either two of my local models.
I also run Agent Zero on my MacBook, I'm thinking of setting it up on my Fedora workstation so it can run 24 x 7, but I haven't got round to that yet.
I'm still deciding whether to push the button to buy a good inference machine, but I am undecided, because I suspect I will end up training, so would buy the Nvidia DGX Spark. But if I just want an inference machine, then the current frame.work desktop machine would be a good candidate, although the Minisforum MS-S1 MAX is perhaps a better choice.
Most of the time TED uses the OpenAI Codex on the rate limited £20 a month plan, but he can now fall back to Minimax on ollama or either two of my local models.
I also run Agent Zero on my MacBook, I'm thinking of setting it up on my Fedora workstation so it can run 24 x 7, but I haven't got round to that yet.
I'm still deciding whether to push the button to buy a good inference machine, but I am undecided, because I suspect I will end up training, so would buy the Nvidia DGX Spark. But if I just want an inference machine, then the current frame.work desktop machine would be a good candidate, although the Minisforum MS-S1 MAX is perhaps a better choice.
1🚀2👀1
