The model is good and runs fine but if you want to be blown away again try Qwen3-30A-A3B-2507. It's 6gb bigger but the response is comparable or better and much

The model is good and runs fine but if you want to be blown away again try Qwen3-30A-A3B-2507. It's 6gb bigger but the response is comparable or better and much faster to run. Gpt-oss-20B gives me 6 tok/sec while Qwen3 gives me 37 tok/sec. Qwen3 is not a reasoning model tho.