Qwen 3.8 27B is excellent, but it defaults to overthinking things
simonwillison.net
Qwen 3.8 27B is excellent, but it defaults to overthinking things
1–10 of 411 posts
Re: Qwen 3.8 27B is excellent, but it defaults to overthinking things
#2The big problem with overthinking on a dense model is obviously the speed hit you take. Going from Qwen 35BA3B to 27B for me is about 7-8x slower (should be ~9x?). This makes me a lot less patient for useless thinking tokens.
I’d want to compare this to the new Muse 30B model which is super terse and has a whole different way of thinking (no “Wait,”) and in my experiments was way more token efficient to the point that the absolute tok / s didn’t really matter.
Re: Qwen 3.8 27B is excellent, but it defaults to overthinking things
#3[flagged]
Re: Qwen 3.8 27B is excellent, but it defaults to overthinking things
#4This is true, but I think it understates the problem. I did a task I've done with a bunch of small models lately (https://github.com/swelljoe/flar/pull/17), and it did an excellent job, the best of any self-hostable model. But, it took eleven (11!) hours on my dual GPU setup. It really chewed on it, and spent a lot of time checking and re-checking. It is by far the slowest model I've used for the task. GPT 5.5 did a similar task in about 20 minutes. Most big models took about an hour or so, and most small models needed a couple of hours (but did a worse job).
Re: Qwen 3.8 27B is excellent, but it defaults to overthinking things
#5The big problem with overthinking on a dense model is obviously the speed hit you take. Going from Qwen 35BA3B to 27B for me is about 7-8x slower (should be ~9x?). This makes me a lot less patient for useless thinking tokens. I’d want to compare this to the new Muse 30B model which is super terse and has a whole different way of thinking (no “Wait,”) and in my experiments was way more token efficient to the point tha…
Comparing with Muse Glimmer is a good idea. I ran the same exact HTML tool generating prompt against both Glimmer 30B and Qwen 3.8 27B. Results:
Qwen: https://gist.github.com/simonw/121ad098860028b2fab603fa12da1... - 17,576 reasoning tokens, produced this HTML result: https://static.simonwillison.net/static/2026/qwen-over-think...
Glimmer: https://gist.github.com/simonw/51e8ddb2ee597a5005fa63bd4927d... 1,021 reasoning tokens, this HTML: https://static.simonwillison.net/static/2026/glimmer-bbox.ht... - ugly but functional.
In both cases paste in the URL https://static.simonwillison.net/static/2026/two-pelicans-on... to see them work.
Both applications work correctly and fulfill the requirements. The Qwen one (which used the default xhigh reasoning setting) is massively over-engineered. The Glimmer one used whatever their default in LM Studio is and I would argue is a tiny bit under-engineered.
Weirdly the Glimmer one doesn't work with images on other domains like https://static.inaturalist.org/photos/714731804/large.jpg - it fails with a CORS error, but you don't need CORS to load images and detect their width and height, and the Qwen one handles that URL just fine.
That's because Glimmer added this unnecessary line:
img.crossOrigin = 'anonymous';Re: Qwen 3.8 27B is excellent, but it defaults to overthinking things
#6This is true, but I think it understates the problem. I did a task I've done with a bunch of small models lately ( https://github.com/swelljoe/flar/pull/17 ), and it did an excellent job, the best of any self-hostable model. But, it took eleven (11!) hours on my dual GPU setup. It really chewed on it, and spent a lot of time checking and re-checking. It is by far the slowest model I've used for the task. GPT 5.5 did…
Was that with the default xhigh reasoning setting? I suggest trying again with reasoning set to low or turned off entirely.
Re: Qwen 3.8 27B is excellent, but it defaults to overthinking things
#7I wonder if this can be fixed with LORAs.
Re: Qwen 3.8 27B is excellent, but it defaults to overthinking things
#8--thinking-budget and --thinking-message is all you need in llamacpp to keep it progressing.
the message can be some combination of tool calling, summarizing, etc. It's overthinking often is a bunch of recursion, so simply stopping t and redirecting is all you need to do.
If someones building a harness for llamacpp, you can set this per message, so it's possible to dynamically control it by watching for the expansion of the thinking traces, and redirecting it.
I use the message to tell it to use subagents, add additional logging and to use opencode's dynamic context pruning.
As such, we'll just whisper here _skill issue_.
Re: Qwen 3.8 27B is excellent, but it defaults to overthinking things
#9The big problem with overthinking on a dense model is obviously the speed hit you take. Going from Qwen 35BA3B to 27B for me is about 7-8x slower (should be ~9x?). This makes me a lot less patient for useless thinking tokens. I’d want to compare this to the new Muse 30B model which is super terse and has a whole different way of thinking (no “Wait,”) and in my experiments was way more token efficient to the point tha…
I love reading Glimmer's "thoughts". Why use many word when few do trick?
Re: Qwen 3.8 27B is excellent, but it defaults to overthinking things
#10I wonder if this can be fixed with LORAs.
I had to fix this on 35B A3B -- I have a proxy that just shuts it down if it gets to 2K thinking tokens and injects something like "We have thought enough, let's begin working." and it almost always finishes the turn then. It rarely needs more than 2K thinking tokens and if it does there is always next turn. I would need to see what 27B is actually doing, but these smaller Qwen models seem prone to this.