This model (35B A3B) is truly great and a clear open source winner right now. I’d love to see it through groq speed.
1 Like
+1 for Qwen3.5
Even something like the Qwen3.5 9B would be great given its multimodal capabilities.
Totally agree! This model is rather small and MoE-based, which guarrantees low latency, but also performs better than gpt-oss-120b and even larger models. It has built-in reasoning, which however can be disabled. So this model could outperform and potentially even replace the current gpt-oss-120b and gpt-oss-20b models.