Now if only they could fix their coding harness so that it's usable for day-to-day use. I paid them for a month to try out Muse Code, but their sandbox never worked across Mac or Linux (both very stock installs, nothing special).
As an aside: I tested out the $20 Cursor CLI subscription, and it created an excellent little graphical app for my kids to play with. The design it came up with was excellent (compared to the crap I got out of Codex for a parallel implementation). I'm going to try to figure out how to get Cursor into my work rotation, at least around design-related artifacts.
Maybe I'm just not looking in the right place, but I cannot find what the API shape looks like. I've even deployed this model via Foundry and it doesn't say what to POST or what to expect back.
How does this compare with Parakeet? I've been using that locally in my projects on an M-series macbook and it's been working great. It's fast and accurate enough for my use cases (meeting transcription, audio transcription for demo videos, etc.)
This definitely seems lighter and faster. How does accuracy compare?
People keep praising Parakeet, but I've found it to be worse than Whisper Large. Yes, it is much, much faster and smaller, but accuracy matters a lot if you are to use dictation regularly and seriously.
I ended up having AI optimize Whisper Large and create a plugin for TypeWhisper, and that's what I use (feeding the results through local Qwen 3.8 running under MTPLX).
When you are running the model locally and trying to do other tasks at the same time, Whisper Large feels too clunky on normal hardware.
I like Parakeet because it's good enough, relatively light, and fairly fast. I'm using this for things like meeting transcription and dictation. Since I'm sending most of the text to an LLM to clean up afterwards, it works well enough.
I'm hoping for something the size of Parakeet (or smaller) but better quality. It feels like with all of the advances in making smaller models better in the LLM space, someone should be able to come up with a lightweight and better quality speech to text model.
I confirm this as my experience as well. Parakeet did not perform well for me (English with heavy accent). Whisper large is much better. But once we get to model of this size look at qwen3-asr (1.7b parameters). I'm quite happy with how it performs (both transcription quality and speed on rtx5080). By the way the model is also available on openrouter and very/very cheap. Performance is reasonable (<1s response), but some requests are delayed (>10s)
Thank you for this suggestion. I am now trying it in 8-bit quantization, and it seems to work very well, at least as well as Whisper Large. It is a little bit faster on my Mac, too.
The real verdict will be after several days of use with dual language dictation, but so far it looks really good!
Whisper large-v3-turbo is still the SOTA non-cloud model? Damn. I'm wanting to move away from it to something smaller/better but based on this comment it seems like I made the right move by sticking with whiser.cpp this whole time. It's been 2 years since large-v3-turbo was released...
I found Parakeet V3 TDT pretty accurate in a meeting app I build. I tried out Moonshine and whisper and found neither of them more accurate or faster than Parakeet. I also didn't like whisper take up more ram/size than parakeet.
Same thing with Nemotron 3.5 ASR. Its just as fast as parakeet and same memory footprint.
It’s really domain dependent I think. If you are doing anything conversational interfacing with less AI-familiar users, latency matters a lot.
If you’re feeding the results into a very smart LLM, it will figure out what you meant (but crucially ONLY if you warn it or tell it to do so, in some cases!). If you’re writing code directly or creating something for public consumption, you can’t tolerate mistakes. If you’re taking notes for yourself you just want it to work cheaply.
If you are ok with the complexity you can run both, and a Meta/Google open model with native audio, and let a smart LLM doctor it up. If performance really matters you can pay for a proprietary model or train one yourself. Until you get to that point, I think it probably doesn't matter much either way. Voice is just too easy to fiddle with
For English I've found Parakeet v2 and Parakeet unified to perform significantly better than v3. For deposition audio in a quiet room and neutral american accents I've hit 96-97% on a number of sessions with v2 and I've seen as low as 83% for heavy accents with overlapping speakers.
Parakeet is the gold standard. With models like moonshine and koroko (TTS model), it’s more about embedding the model in the application itself. If you’re using Parakeet, embedding it in the application is not feasible.
I use parakeet with superwhisper, and I’m making another app that has SST and TTS built in, and I want to use my downloaded parakeet model, but it seems there’s so many different implementations from ONNX to whisper, it’s not easy to use your downloaded models. So models like moonshine and this one allow you to just embed it into your application simply. It might not be as good as parakeet, but it gets you 80% of the way there.
The cost / performance chart shows that in almost all configurations, it looks worse than Opus. Why would you use Sonnet 5.5 on xhigh if you would get better results (higher score, cheaper cost) on Opus 5.5 high?
Is there a good use case? This isn't like Luna where it's much cheaper/effective just to use Luna in certain situations.
Per the charts, there is largely no point to using Sonnet 5.5 at high+ as opus low generally will give similar performance at similar or lower cost.
But Sonnet 5.5 at medium and below gives you a cheaper option at a performance worse than the lowest thinking Opus (low), which may be viable for "low intelligence" use cases.
It appears, at least from a quick look, to be noticeably faster than Opus. If true, and you don't need xhigh/max reasoning for your use case (like a well-defined set of code changes), Sonnet might get the job done much more quickly.
With that said, at that point, I'd probably use something like DeepSeek V4.1 Flash, which is way faster and significantly cheaper, and probably not noticeably dumber for most use cases.
That makes sense. I'm interested in seeing where Haiku 5.5 comes in then when it gets released. It feels like the low intelligence / fast niche will be covered there.
My guess is that haiku will be a mid release. They don’t seem to care to compete for the low end. Something akin to gpt6 sol level intelligence at $1 / $5 pricing. Then not release an update for 6+ months.
It depends entirely on its capabilities. If it is significantly smarter than Luna, which frankly is quite likely, then a lot of people won't mind paying more.
Or if its luna at 1000 tok/sec. Speed is what most of my peers care most about these days since less intelligent models can do most grunt work just fine.
There's a sort of magical thinking needed to answer a question like that. You might say it comes down to "feel" of the model; i.e., the indefinable differences in the way that they speak to the user and approach problem solving. Perhaps Opus is suited for tasks that tackle new ground, while Sonnet might be better at tasks that are more grounded in the code.
Ultimately it's slightly ridiculous to define model capability on a single axis. It's like a standardized test. Sure, you can line people up by their ACT score, but that doesn't mean a doctor and a brilliant artist who both do well on the ACT have an identical intelligence or approach to life. It just can't be captured.
just shows you how little control of output these labs actually have. They are training two models that kind of ended being the same so whatever they were doing specifically didnt make much difference.
The cost / performance chart shows that in almost all configurations, it looks worse than Opus. Why would you use Sonnet 5.5 on xhigh if you would get better results (higher score, cheaper cost) on Opus 5.5 high?
This screams to be that Sol vs Terra model problem that OpenAI had. On paper half the price, in actual usage the price gap was so close for less good results, that everybody just spammed Sol.
I'm honestly not sure where they're getting their 30% numbers from at all. In every single chart that they chose to display except for one, it costs similar or more than Sonnet 5, while also being comparable in price to Opus.
Maybe it's buried within their system card but I think that this would be one of the first things they'd want to show in the announcement article and they fail to do so.
I really don't know who does Anthropic's marketing but they always seem to a pretty terrible job in their announcements from my perspective.
Are people using MiMo models as their daily drivers in a company setting? If so, how? I know they're available over OpenCode and directly from Xiaomi, but those are not great options. OpenCode Go straight up doesn't give any guarantees about training on your data, and Xiaomi says that they won't but it's unclear.
With some models you can find hosting companies based in the EU or US, but then you don't know how they're quantizing the models, so you're not sure about the actual output quality.
How are people actually using this? Or are people just experimenting with side projects?
I only use it for personal projects, but their european Token Plan [1] has ZDR and is hosted in Amsterdam. The DC also likely runs on 100% solar wind/power, with heat recovery systems used to heat nearby university buildings.
Hmm I may be mistaken on that now, I remember it being in the user agreement. There are third party routers that explicitly offer ZDR, though API costs will be slightly more expensive.
Yep am waiting for some of my whitelisted openrouter providers to pick it up (so far none offer it but hopefully a few will in the next couple of weeks)
You can check out Muse Spark 1.3 by using OpenCode (https://opencode.ai/ - open-source AI / coding harness). There's a terminal version and a GUI / desktop version. Good luck!
If the benchmarks are real and reflect actual use, then this is an insane model. This 300B model outperforms the previous DS4 Pro preview model (1.8T params), and it looks like it outperforms GPT 5.6 Luna too. And it's still cheaper than Luna, even with the price decrease.
For reference, that puts it on the third spot behind GPT 5.5 and Fable 5. For some reason GPT 5.6 Sol is not showing in the leaderboard. If it did, then DS4 Flash was number four.
The thing is, even if Luna is better in DeepSWE and has the 80% discount. DeepSeek is still cheaper.
--------------
DeepSeek V4 | Flash GPT-5.6 Luna (New)
--------------
Input (Cache Hit) $0.0028 $0.02
Input (Cache Miss) $0.14 $0.20
Output $0.28 $1.20
--------------
Both Luna and Flash are heavy on the reasoning > output. And the cache hitrate + prices also matter.
Reality is, you can not go wrong with Luna or Flash at those prices. And remember, DeepSeek V4 Pro is still in the rafters, what is ironically closer to Luna's new price.
Now if only they could fix their coding harness so that it's usable for day-to-day use. I paid them for a month to try out Muse Code, but their sandbox never worked across Mac or Linux (both very stock installs, nothing special).
reply