Qwen3.6-35B-A3B：Now Open-Source (opens in new tab)

(twitter.com)

3 pointsalexellisuk24d ago1 comments

1 comments

The team say it beats 3.5 27B in numerous benchmarks. But it's still a MoE with 3B active per token, so we'll see!

I hope there's a 3.6 27B coming in short order, or even a 31B.

Gemma 4 looked great, but I've not got it working well with Claude code (Qwen just works) - Gemma 4 _does_ work with opencode though.

j / k navigate · click thread line to collapse

The team say it beats 3.5 27B in numerous benchmarks. But it's still a MoE with 3B active per token, so we'll see!

I hope there's a 3.6 27B coming in short order, or even a 31B.

Gemma 4 looked great, but I've not got it working well with Claude code (Qwen just works) - Gemma 4 _does_ work with opencode though.

j / k navigate · click thread line to collapse