TQ1 quant?
#6
by sergeysi - opened
Very thankful for your work.
Could you please make UD-TQ1 quant? Would be very helpful to run on iGPU potatoes.
what even is the point of running models at such an unusable quant? why not run something way smaller but usable?
I'd argue it is usable. For example Qwen3-Coder-Next-UD-TQ1_0.gguf is quite usable (probably thanks to Unsloth Dynamic).
Same, UD-TQ1_0 of other qwen models (next and 3.5) have been much better than you'd expect.