Comment by vrc
9 hours ago
Out of curiosity and semi unrelated — why do so many of these projects with customized encoder-decoder setups use earlier Qwen versions like 2.5 and 3 and not the smallest 3.5? Purely the few 100m params, or something else in the latter’s arch or pretraining?
In my experience if you tell Claude to port LLM-like stuff without explicit steering for versioning, it will default to the most popular thing for this in its training window to reduce errors. 3.5 is outside its training data.
I came across this recently. I was scanning for tiny models from HF using their search API. The script was generated by an agent. When I ran it, Qwen 3.5 did not make it at the top. Turns out, models generally prefer older content (training) but that the scanner also did not give any importance to recency.
My guess would be qwen 2.5 predates linear attention which would be more complex to use.
Anecdotally qwen2.5-14b runs a lot faster on my mini than the small 3 models ¯\_(ツ)_/¯