OpenAI's GPT-OSS models benchmarks worse than DeepSeek R1 and Qwen3 235B

Not entirely unexpected given that these models have 2x - 5x more parameters than the 120b gpt-oss model.

It’s impressive their performance gets so close given the size!