Need help to buy a new computer, which coding model is the best atm?
I need to run local models eventually to start working on harness optimizations, adding local power to my subscriptions when possible
The thing is, I have no idea which model is the best for coding locally at the moment, have seen comments on Minimax 2.7, Kimi, GLM, Deepseek, Qwen, but they all differ on different benchmarks and need some guidance from experience if possible to see how much VRAM I need to actually run them locally