Ollama with PARAMETER think support + CUDA 3.5/3.7 (Tesla K80) + ROCm gfx803 (RX 580) legacy GPU support
amd cuda nvidia multi-gpu rocm rx-570 local-llm llm-inference ollama rx-580 modelfile deepseek-r1 ollama-fork tesla-k80 cuda-compute-37 cuda-compute-35 legacy-gpu rocm-gfx803 rx-480 thinking-mode
-
Updated
Feb 6, 2026 - Go