There is an old rule of thumb that says the quality of an MoE is equivalent to a dense model with the geometric mean of its total and active parameters. So, for example, the Qwen3.6 would be equivalent to a dense model with approximately sqrt(35×3) ≈ 10.25B parameters.
replyBoth MoEs and dense models are always getting better, so I don't think this comparison is meaningful across generations. But still for a first approximation, this tends to hold (you wouldn't expect a lot from a 10B model in coding yet).
> ollama's model index typically only distinguishes variants by their size
replyDon't use ollama. The entire project is just a series of stupid decisions like this.
What's the sota? The classic vLLM/llama.cpp? LM Studio? Unsloth studio any good?
replyHow they have any credibility after they told everyone they could run Deepseek R1 on their laptop (by giving the Qwen 7B distill the `deepseek` moniker)...
replyAnd last time I have checked, you still can't run rerankers with it, yet you can download the models. See issue 3368, 2years old now.
replyI've been pretty impressed with Laguna. I downloaded their coding agent and have used it for a task here and there (the larger variant). Obviously it's nothing like a frontier LLM, but it surprised me with how good it was. And I think the model personality and way it talks is pretty pleasant
replyI’m confused about the naming suggestion. Seems like the AxB bit differentiates pretty clearly, no?
replyOnly if you know what it means. To many people these suffixes are gibberish. Also ollama doesn't actually show that suffix in its model index. Don't say "don't use ollama", you sound like a Linux user telling people not to use Mac. (I say this as a Linux user who despises Apple products myself.)
replyOllama uses "consumer UI guidelines", it's the second worst AI tool to use for anything more than writing fanfic.
replyNow I have to ask – which is the worst?
replyGood luck writing fanfics with primitive top-p and top-k samplers.
replyWhat’s good for data extraction? I have a hard time getting models to just pull names and titles from a blob of text.
replyBuild/vibe-code your own benchmark and find out, the work pays off and lets you try many models.
replyAlso try fine-tuning small models and see how they perform.
If you haven't yet, you might want to try gemma-4-E4B-it-qat or gemma-4-12B-it-qat with Structured Outputs. My main use case is tagging photos and generating headlines.
replyLaguna XS is MoE, however.
replyOh!
replyI think I missed that and assumed it wasn't because it performed similarly to the dense models. Interesting!
As an aside, the bigger S 2.1 runs faster than the dense Qwen 3.6 and Gemma 4 models on the same hardware (assuming the same hardware is big enough to run it) and definitely feels smarter, and more capable of long tasks, but it doesn't feel as heavily tuned for code as Qwen 3.6.
reply