This thing is crazy fast in comparison to base Qwen 3.8 27B!
The video I'm attaching is having it on xhigh reasoning so a lot of prose in thinking but it's still crazy fast - I'm really impressed!
Omni A/V agents just hit $0.15/$0.47 per MTok — Switch or Skip?
Qwen3.8-Omni-Flash (qwen3.8-omni-flash): text/image/audio/video → text, 1M ctx. International Model Studio: $0.15/$0.47 per MTok · $0.016 cache hit. Qwen: A/V near Gemini 3.8 Flash; >98% lower audio-input $/hr vs
The timeline spent two days saying bonsai 2 cannot build.
Here is 1 hour 24 minutes of it building, one shot, from one paragraph, on an rtx 3060 12gb, sped to 8x so you can watch the whole thing. what you are watching is a 5.9gb ternary compression of qwen 3.8 27b, served
Thoroughly unimpressed with Ternary-Bonsai 2
Its commendable what they are trying to achieve but this is nowhere near the quality of Qwen 3.8 27b let alone 98.2% vs BF16
You have reached the end of the archive
All of qwen38