But I have to say, I quite like the way Muse Glimmer thinks and talks. It's a cocky bastard in tone, but it's quite good, and its thinking traces are relatively terse.
AI companies release models when they are ready - not on a set schedule. The steps are required to produce a release candidate, so a company can choose to bless the RC with the best metrics at any time - or alternatively continue (post/)training newer RCs if they feel they can do better, and they have time. When a model is ready for release is subjective, and can take into consideration what the competition is doing.
An obvious counter-example to schedules driving releases is the still unreleased Gemini 3.6 Pro.
Just recently, Minimax H3 released as open weights on the eve of Seedance 2.5 global availability. It's not as good, but it's good enough and it's completely open.
Flux 3, which is nowhere near as good as either, suddenly announced their release once news of these other two became public. They knew if they waited they'd be ignored. It didn't really help them much, unfortunately.
The LLM releases are even more rivalrous.
And don't forget all of the competing launches planned before Google IO or major release events.
Companies like to eat into the news and press cycle of their rivals.
Minimax H3 can run exceptionally fast (10 minutes for a 15 second 0.5mp video and that's stock cuda 13), works on 16 GB VRAM GPUs, etc. If Flux3 is anything like Flux2, it’s going to require an absolute monster truck of a machine and still run significantly slower. Even if it’s a better model, that won’t matter as much if nobody releases any LoRAs or fine-tunes for it.
Not to mention BFL licensing often feels deceptively confusing and restrictive.
Seems a bit premature of a statement lol
Even if I did, we’re talking barely a decade
The whole class of what can reasonably fit in a single GPU is an interesting category of LLM, and based on the results I've seen from 3.6 35B A3B and 27B versus what existed a year prior, it seems there's a lot of room for advancement.
As people have called out, Glimmer appears to be a trade-off rather than a clear winner.
And from what I've been reading, no one is expecting Qwen 3.8's model in this space to be a clear winner, but just slightly and marginally better.
That's a little concerning as DeepSeek v4 Flash proved at it larger sizes there's a ton of room left to compress knowledge.
If we don't see something that's substantially better in the ~30B param space soon - it would appear we might've saturated that size with knowledge.
I wouldn't be quite so pessimistic. We may have saturated the current approach, but I think there's a lot still left in terms of compression, attention, active parameters, caching etc. etc.
For contrast, the Newcomen steam engine (widely considered the first commercially useful engine) was used for over 60 years before the next major improvements. Now, 300 years later, we’re still finding ways to significantly improve heat engines.
Off-topic, but I stumbled upon the first Newcomen engine imported into Australia in a museum in Sydney and I was unexpectedly charmed (not an Engine Guy). It's large, but nothing like the awe of "mega-engineering", it's crude, but it clearly has such amazing utility (when compared to a reality without it) and it changed the world
Qwen 3.6 27B was already a massive gift to smaller homelabs around the world; anything more is just a delightful surprise.
I suspect this is the only use-case I would consider...and I don't really have a use-case for "creative writing" that I would delegate to an LLM. I suppose for dialogue generation in games?
But yes, hard agree. Why on Earth would you ever want to write code with a model that is supposedly "jailbroken"? So it can put great backdoors into everything it touches? Pass.
I use it to write short sci-fi stories. Life is not only about being an SDE.
> Why on Earth would you ever want to write code with a model that is supposedly "jailbroken"?
I need a model which decent enough at coding to write boilerplate for me, yet I do not need to unload in case I want to write an article or enguage in my writing hobby.Exactly why Gemma 4 are so popular.
> So it can put great backdoors into everything it touches?
What a strange take...
It seems to know tooling and coding.
</div> is four Gemma4 tokens, but one Qwen3.6 token.
In that file you can see all tokens and their corresponding numeric codes.
Considering that Meta distills Qwen[1] (and should!), it'd be hilarious if Muse loses the head-to-head; the "distillation attack!!1!" people claimed distillation on release n-1 is enough to match the intelligence of the latest version.
1. They wrote a paper about it
Each turn is about 45-60 seconds to generate all of the various responses. The GM and director have reasoning on, and the NPCs/Location/Narrator do not.
It's a fairly good "engine" for that. I'm not sure how a denser Qwen would do here regarding speed.
I'm not sure what all is needed to make that work for people.
Back in 2023 I started my own C# LLM library for doing tool calls and structured output, and over the years it has morphed bigger and bigger, and that is the backbone of almost all of my LLM-based projects.
I've never released it, but its easy to understand, and simple to add your own tools:
[AIDescription("Get current weather for a location")]
static string GetWeather(
[AIDescription("The city name")] string city,
[AIDescription("The country name")] string country,
[AIDescription("Temperature unit", ["C", "F"])] string unit = "C")
{
// make some API call to a weather API and return a string to the LLM
return $"The weather in {city}, {country} is 22°{unit} and sunny";
}
var chat = client.StartConversation("You are a helpful assistant with access to weather data.");
var response = await chat.SendAsync<string>("What's the weather in London?", GetWeather);
If you’re in the C# ecosystem you could consider converting your APIs to MCP format tools using the MCP SDK.
https://devblogs.microsoft.com/dotnet/build-a-model-context-...
And then leveraging Microsoft Agent Framework for the client and orchestration side of things:
My issue currently is KV Cache, because I can't keep enough parallel caches running (4 is where I'm at), so TTFT (is that the initialism?) can be long when I have a particularly large scene (basically more than 2 NPCs).
But my harness does let me offload to any OpenAI compatible endpoint, I just prefer local cuz free.
The Qwen models have a habit of going into thought loops where they go in circles for a while.
Try a system prompt requiring it to think in Mandarin, while still delivering the response in the user’s language.
PS: Switching the thinking process from Chinese to English can also significantly circumvent certain self-censorship mechanisms built into the model.
Once you get the agent and message consistent,itll keep moving.
Edit: arXiv:2505.09388 Qwen3 Technical Report
Huh, well... no? Gemma A4B and Qwen A3B are quite popular in fact. I'm sure 3.8 35B A3B will outperform 3.6 27B by all metrics
Qwen3.6 is a definitive, significant downgrade from Qwen3.5 for creative writing and prose for example. Yes, it's better at agentic and coding, but it regresses in many non-coding areas compared to Qwen3.5.
Of course, I do expect the 3.8 ones to perform better for agentic coding.
Tech tends to get boring when you judge current products against the hypothetical capabilities of unannounced products that may never ship. It's like comparing Nikon cameras against Canon camera rumours, or comparing iPhones against unannounced and therefore largely imaginary Samsungs.
- If they do a Qwen 3.8 35B A3B (and I hope they do because I love the 3.6 version)
- and if it beats 3.6 27B by all metrics
… then the local open weights world will be a better place.
But they have said nothing about it and they dropped several weight classes for 3.6, so who is to say they won't drop the 35B? And even if they don't, this is a tall order; why would the MoE tradeoffs no longer be apparent? (Again, I really like both the Qwen and Gemma MoEs)
FWIW I am enjoying testing Muse Glimmer — it's really quite impressive on chat, has nice terse and even amusing thinking traces, a bit of brass to it, and I'm hoping it will be good on agentic stuff.
Surprising that Meta don't host this model, even as rate-limited free-tier.
> open weight version of Muse Spark 1.2
Wait. Is this "version" different from what Meta serves?
UPD. was wrong on smaller, it's actually much larger
- muse glimmer: 15.9GB
- qwen 3.6 27B: 17.6GB
My video card is so close to its limit that these GB thresholds are mattering too much for me :D
UPD, NVM, got misled by comments here. It is actually almost 60 GB so much larger
Meta actually relesed official 4 bit quants in 17GB, but I haven't seen any indication that training was quant-aware, so the quants are not going to have same performance. 3.6 27B has official FP8 quant that AFAIR was trained with quantization awareness.
The best example is last year's gpt-oss which was released prequantized in mxfp4 so 20B parameter model was under 14GB and 120B was under 70GB right away.
readme on huggingface says they've benchmarked the quants -- for 17GB quant reported 1% avg loss across 15 benchmarks (sadly no breakdown).
I assume that's strong enough signal for QAT. Not just first party quants, but they cared to monitor degradation.
That's exactly the point. We know short context knowledge stuff does not regress with quantization. But I expect agentic intelligence to suffer greatly.
If I were to pick one bench, I would like to compare quants on TerminalBench Hard. But then Glimmer already loses to 3.6 27B on it by a large margin.