DeepSeek R1 32b is really good BUT...

I do a deep dive into DeepSeek R1 Qwen2.5 Distilled 32B parameter model. I benchmark a few models on my 7900xtx, and then showcase about a dozen results from directly using the DeepSeek R1 32B on my local computer. Overall it did surprisingly well.

I did find a few oddities that I call out in the video, such as talk about China in the chain of thought.

Let me know if you like seeing local LLM benchmarks, trying to gauge interest in that.

#deepseek #deepseekr1 #localllm #ai