← Back to context

Comment by syntaxing

2 hours ago

I am so excited for Qwen 3.8 27B. It’s a shame how slow prefill (~3-400) is on a strix halo but it’s such a good model for agentic tasks.

Prefill is survivable if you cache well. But what kills me is the context. Qwen 27 needs a ton of room for KV Cache. I guess not an issue on a 128 GB Halo or Spark, but if you are running of consumer/prosumer GPUs it's miserable to be compacting every 120k tokens.

What type of agentic tasks are you using it for (eg how complex)?

  • For personal stuff, I use it with AnythingLLM. It replaced any Google search for me. For coding, I run opencode though I have been debating switching to Pi. I would argue it’s at Sonnet 3 level.

I find that 35B-A3B is much easier to run on my M4 Max (both prefill and generation)

  • It's well known 35b is much faster (on any hardware) and quite a bit dumber

    • This really very much depends on how you are using it, I think. If you intend to leave it to solve long context problems and write whole prototypes, the 27B is going to be much better.

      But if you are sort of pair-programming with the model, the speed obviously matters and I think then the 35B is acceptably smart, and when it's wrong it'll be wrong much more quickly. It seems very good on SQL and PHP, and I assume on typical JS and Python.

      I would rather work that way, so I hope they do produce a small MoE model.

How are you running it on a Strix Halo? The weights aren't out yet, are they?

  • I interpret @syntaxing as meaning they are looking forward to running Qwen3.8-27B, but are frustrated by prefill times with other models, such as Qwen3.6-27B.

  • I meant Qwen3.6. Unsloth supposedly has early preview of the model and the VRAM requirement is the same so most people expect similar model size and type.