Category narrowing to this degree really has diluted the term "SOTA" quite a lot
You mean it's state of the art for this subset of benchmarks you're targeting? No? Well surely it's state of the art for models of a certain size category? Huh, what about models that are available for local inference? O, state of the art for open models of a certain size made in the US specifically. Uh. Neat!
Playing games like this also makes it very easy to accidentally say things that aren't true, like in the headline of the article...
> Introducing Ripple-65B Preview, a Frontier Open-Weight Model
It's not "frontier", it's not "frontier" even in the category of open-weight models (though usually that phrasing would mean frontier over all, and also open weight)... it's only "frontier" in the small sub-category of US made open-weight models.
It is technically frontier, if by "frontier" we mean the Pareto frontier.
Being at the Pareto frontier means that you can't get something better on one aspect without compromising on some other aspect. The trick is that it is easy to be at the Pareto frontier by having enough dimensions.
Here, we have 4: score, number of parameters, origin (US) and open-weight. You can get a better score or a lower number of parameters, but you have to compromise on at least one of the other aspects. That's a pretty weak frontier if you ask me.
“Origin” isn’t part of the Pareto frontier because it’s not a quantitative dimension.
(You could say that “ability to be trusted by US government/military users” is quantifiable, but I would argue that by that metric, this model falls behind non-US-originated open-weight models that have more eyes on them!)
Discrete dimensions can exist in Pareto considerations. Not that I'm too fond of this game, but given that "Origin" is definitely part of the calculus of model deployment, it makes sense to consider it in a Pareto calculation.
It's just a different pareto frontier because you're filtering out items that don't fit your criteria. "Model was made by Anthropic" would be an equivalent dimension. You would never plot this, you're just reducing the subset of models you're comparing.
By that logic, any model is a frontier model. My GPT-2 clone is a frontier model when you compare it against models that were trained on my homelab.
AFAIK the currently biggest US open-weight model is Inkling, which they include in the comparison, so at least their claim is not limited to a size category. Still they score 0.42 percentage points higher than the second-best model on their chosen benchmark, so it's hardly a meaningful lead.
Why should US or China matter if its open weights? Its going to matter a whole lot more whether you are running it locally or where your provider is surely.
[Edit i think the preferred combination for those who actually want some data security at the moment and arent at the scale of self hosting llms is chinese open weight models served in germany or France where theres some better data privacy/security]
For sure, but unfortunately, there's a non-zero chance the state department at some point makes the "US" part of this mandatory for a bunch of companies.
Terms can have multiple definitions and language is descriptive not prescriptive. If you intend to communicate that you mean "Pareto frontier" in the current context you need to include the word "pareto." If you don't then "frontier" will be understood as referring to mythos/astra/opus 5.5 tier models in intelligence (max capability regardless of tradeoffs). It doesn't matter that the second definition in the dictionary is the one you want to convey (charitably assuming you're not trying to mislead given the value of the hype of being able to call your model a frontier model) if your audience understands what you're saying to mean the first definiton.
Irrelevant. A term can be well-defined and used correctly while also being diluted. In this case, I am pointing out that it's diluted by narrowing the category in which the claim is being made. I have no doubt that in this narrowed category, the term is being used correctly. This has no bearing on the fact that by narrowing the category in that way, this claim is made remarkably less impressive or interesting to many people hearing it, including myself
It's great to see another competitive open weight model. Congratulations!
But as is, it's probably hard to become popular amongst hobbyists and local model enthusiasts like Qwen's 27B class models or MOE models because it's a 65B dense model and inference will be prohibitively slow on consumer grade hardware, and seems like the goal/mission of the company behind this model is to automate chip design.
So if I have to assume and connect the dots, this is for the companies who want to use AI in air-gapped environments for chip design so that they protect their IPs (not leaking to third party AI labs/hostile parties) and can afford lots of server grade inference hardware for a quick unlock with the big dense model.
> it's a 65B dense model and inference will be prohibitively slow on consumer grade hardware
Times are changing and with MTP and DFLASH, dense models are back in style. For qwen 3.8 27b, you can get great token generation rates with MTP and for quality, it's much better than a MoE of comparable size.
Maybe is a bit more niche, but I'm pretty sure there is plenty of people who have enough RAM to run dense models like this one, I don't know how slow is slow but people are willing to sacrifice speed if it means they get a good model running locally.
Look at "SC117/Qwen3.8-Flash-Next-GSQ-RCO-abliterated-GGUF" on HuggingFace, It says it has 3.5M downloads this month and its offering a model that is 2-bit and 3-bit quantized (66GB - 83.9GB).
If this gets quantized to 4-bit (Q4_0) which usually there is very little loss in intelligence compared to the base model, then the model will probably be around 40GB, and looking that people are downloading a model that is at least 20GB bigger, then it might allow even more people to run a bigger model.
But it will only be popular among hobbyist if its as good as it claims to be.
Qwen 3.8 Flash Next is a mixture of experts model though. It has 6B active parameters per token, so you can cache large parts of the model in RAM and move it in and out of VRAM as needed. If you do that with a dense model it gets much slower.
Experts are not typically moved between VRAM and system RAM during inference, as it would be too slow.
What happens instead is that the "dense" parts of the model like e.g. the router are put into VRAM and the remainder of VRAM is occupied by the KV cache and whatever experts still fit.
The rest - usually just the experts - reside in RAM, but it's bo big deal.
Additionally, some models utilise an n-gram table, of which only one row is read per token, so it can actually be stored on disk without issue. Example:
> Experts are not typically moved between VRAM and system RAM during inference, as it would be too slow.
For some definition of "typical", sure, but it depends on what performance trade someone is willing to accept. Strata LLM is geared around streaming weights as needed. It lets me run DeepSeek-4.1-Flash 1031, heavily quantized (ISTA-DASLab's 2.25 bpw quant), on a single Radeon R9700 plus 256 GB host RAM. 500 tps prefill and 30 tps decode isn't great, but it's a large fraction of what I get with Qwen3.8-27B quantized to fit entirely in VRAM for short prompts, it's faster past about 80k context, and DS4.1F can use a 1M-token context. It takes forever to fill that, or even to refill if the KV cache doesn't match, but it works.
Something doesn't seem right. Vanilla llama.cpp can hit 50 t/s on an R9700. Q6 (albeit with limited context of 64k, more if kv cache is quantized) fits entirely on the card.
I hear what you're saying - there are people who will use it even at 0.1 tokens/s and run tasks over night or over multiple days if it is good. but Qwen3.8 Flash Next is a MoE model (6B active) so the memory bandwidth/compute required to calculate each token is very different from a 65B (meaning, 65B all active) dense model!
Is it competitive overall or just competitive when compared to "U.S. Open-Weight models that are otherwise pretty decent at agentic coding"? The latter does not imply the former.
What'll be interesting is how it performs relative to Qwen3.8 27b when quantised to the same size. Which, of course, is a comparison they don't want to do.
Well, obviously. These newer American models aren’t competing with Qwen yet. They are competing for American businesses and organizations that can’t/won’t use Chinese models.
They are playing catch-up, but I’m happy to see new players to the game, and it will be interesting to watch how they develop
What's funnier is that after cherry picking models for comparison, they further cherry picked Inking for size comparison instead of better and smaller Beam 1.0
The bar chart also shows Nemotron 3 Ultra instead of smaller and better Nemotron 3.5 Super.
I think it's because the primary users of this model won't even look at "those other open weight models" as options. This is the right chart to present to those who would consider this model. (e.g. US Gov't agencies, chip design industry, etc etc)
US Gov won't let Chinese models be hosted on any internal infra, not even supposed to use them for local AI. So there is absolutely a market for US-made or non-Chinese models there.
Because Chinese models heavily rely on distillation and regularly behave strangely in public benchmarks (where they score much better than in private evaluations), even if many world class researchers are Chinese. It's difficult to quantify to what extend labs like Moonshot would be competitive if they could not distill from OpenAI, now that Anthropic mostly blocks it
There are several situations where some of us are explicitly prohibited from using models from companies not under the jurisdiction of the United States. It doesn't mean these models are competitive when looked at from a global view but it is also not an irrelevant differentiation.
Speaking as someone who has High-Memory Mac Studios, it is great that there is finally a model that will run at usable speeds with high quality. In my experience, the sweet spot for Mac Clusters is around 200GB of weights, which this is far under. Congratulations to the recursive team for building an open model that will actually run!
I understand the desire to look the best in some self defined category. But not showing any comparison against the actual leaders in open weight renders all of the benchmarking completely information-less for the actual end users (who care about real performance comparison against the real field).
Not comparing to Qwen3.8 27B gives cherry-picked benchmark comparison energy. Still, Ripple-65B outscores Qwen3.8 27B, according to the reported benchmarks.
Showing a terminal bench 2.1 score at the top is very misleading when terminalbench 4.0 is the current metric as a very quick glance for anything we care about for 30B to 180B size possible to locally-host models.
So the model is "meh" at best, but the cost of the training is interesting. So we can extrapolate that in 3 years the cost will be below $5m for a model like this.
It's very well within the capabilities of any reasonable organization, maybe even crowd-funded.
What is this thing where model comparisons blatantly pick other crap models and show themselves at the top left (cheap and effective)? Put all the open models, or put all the cheap models, or whatever, but just cherry picking is egregious.
Category narrowing to this degree really has diluted the term "SOTA" quite a lot
You mean it's state of the art for this subset of benchmarks you're targeting? No? Well surely it's state of the art for models of a certain size category? Huh, what about models that are available for local inference? O, state of the art for open models of a certain size made in the US specifically. Uh. Neat!
Playing games like this also makes it very easy to accidentally say things that aren't true, like in the headline of the article...
> Introducing Ripple-65B Preview, a Frontier Open-Weight Model
It's not "frontier", it's not "frontier" even in the category of open-weight models (though usually that phrasing would mean frontier over all, and also open weight)... it's only "frontier" in the small sub-category of US made open-weight models.
It is technically frontier, if by "frontier" we mean the Pareto frontier.
Being at the Pareto frontier means that you can't get something better on one aspect without compromising on some other aspect. The trick is that it is easy to be at the Pareto frontier by having enough dimensions.
Here, we have 4: score, number of parameters, origin (US) and open-weight. You can get a better score or a lower number of parameters, but you have to compromise on at least one of the other aspects. That's a pretty weak frontier if you ask me.
“Origin” isn’t part of the Pareto frontier because it’s not a quantitative dimension.
(You could say that “ability to be trusted by US government/military users” is quantifiable, but I would argue that by that metric, this model falls behind non-US-originated open-weight models that have more eyes on them!)
Discrete dimensions can exist in Pareto considerations. Not that I'm too fond of this game, but given that "Origin" is definitely part of the calculus of model deployment, it makes sense to consider it in a Pareto calculation.
It's just a different pareto frontier because you're filtering out items that don't fit your criteria. "Model was made by Anthropic" would be an equivalent dimension. You would never plot this, you're just reducing the subset of models you're comparing.
By that logic, any model is a frontier model. My GPT-2 clone is a frontier model when you compare it against models that were trained on my homelab.
Ahh I run into this issue all the time with ai when applying any sort of dimensional reasoning abstractly it always implicits continuous
AFAIK the currently biggest US open-weight model is Inkling, which they include in the comparison, so at least their claim is not limited to a size category. Still they score 0.42 percentage points higher than the second-best model on their chosen benchmark, so it's hardly a meaningful lead.
Fourth most popular comedy folk rock duo
..in New Zealand!
The US bit got me as well.
I guess it makes sense - there is a segmentation where users may want to run on domestic "frontier LLMs."
Why should US or China matter if its open weights? Its going to matter a whole lot more whether you are running it locally or where your provider is surely.
[Edit i think the preferred combination for those who actually want some data security at the moment and arent at the scale of self hosting llms is chinese open weight models served in germany or France where theres some better data privacy/security]
For sure, but unfortunately, there's a non-zero chance the state department at some point makes the "US" part of this mandatory for a bunch of companies.
No, the term is well defined. It means being on the Pareto curve of performance vs size, such as: https://tinyurl.com/26sw35fw
Terms can have multiple definitions and language is descriptive not prescriptive. If you intend to communicate that you mean "Pareto frontier" in the current context you need to include the word "pareto." If you don't then "frontier" will be understood as referring to mythos/astra/opus 5.5 tier models in intelligence (max capability regardless of tradeoffs). It doesn't matter that the second definition in the dictionary is the one you want to convey (charitably assuming you're not trying to mislead given the value of the hype of being able to call your model a frontier model) if your audience understands what you're saying to mean the first definiton.
Irrelevant. A term can be well-defined and used correctly while also being diluted. In this case, I am pointing out that it's diluted by narrowing the category in which the claim is being made. I have no doubt that in this narrowed category, the term is being used correctly. This has no bearing on the fact that by narrowing the category in that way, this claim is made remarkably less impressive or interesting to many people hearing it, including myself
Is it really on the pareto curve if you get to define arbitrary conditions for what other models are plotted on the curve?
well yes but the y axis heavily depends on the benchmark you choose
By...whom? An ISO standard? Highly cited scientific literature?
A tinyurl pic linked by randoms over the internet?
State of the AI-Apocalypse.
It means ot was released recently.
It was already meaningless when US labs used it.
It means new.
It's great to see another competitive open weight model. Congratulations!
But as is, it's probably hard to become popular amongst hobbyists and local model enthusiasts like Qwen's 27B class models or MOE models because it's a 65B dense model and inference will be prohibitively slow on consumer grade hardware, and seems like the goal/mission of the company behind this model is to automate chip design.
So if I have to assume and connect the dots, this is for the companies who want to use AI in air-gapped environments for chip design so that they protect their IPs (not leaking to third party AI labs/hostile parties) and can afford lots of server grade inference hardware for a quick unlock with the big dense model.
Neat!
> it's a 65B dense model and inference will be prohibitively slow on consumer grade hardware
Times are changing and with MTP and DFLASH, dense models are back in style. For qwen 3.8 27b, you can get great token generation rates with MTP and for quality, it's much better than a MoE of comparable size.
Maybe is a bit more niche, but I'm pretty sure there is plenty of people who have enough RAM to run dense models like this one, I don't know how slow is slow but people are willing to sacrifice speed if it means they get a good model running locally.
Look at "SC117/Qwen3.8-Flash-Next-GSQ-RCO-abliterated-GGUF" on HuggingFace, It says it has 3.5M downloads this month and its offering a model that is 2-bit and 3-bit quantized (66GB - 83.9GB).
If this gets quantized to 4-bit (Q4_0) which usually there is very little loss in intelligence compared to the base model, then the model will probably be around 40GB, and looking that people are downloading a model that is at least 20GB bigger, then it might allow even more people to run a bigger model.
But it will only be popular among hobbyist if its as good as it claims to be.
Qwen 3.8 Flash Next is a mixture of experts model though. It has 6B active parameters per token, so you can cache large parts of the model in RAM and move it in and out of VRAM as needed. If you do that with a dense model it gets much slower.
Experts are not typically moved between VRAM and system RAM during inference, as it would be too slow.
What happens instead is that the "dense" parts of the model like e.g. the router are put into VRAM and the remainder of VRAM is occupied by the KV cache and whatever experts still fit.
The rest - usually just the experts - reside in RAM, but it's bo big deal.
Additionally, some models utilise an n-gram table, of which only one row is read per token, so it can actually be stored on disk without issue. Example:
https://huggingface.co/ISTA-DASLab/Qwen3.8-Flash-Next-GSQ-RC...
> Experts are not typically moved between VRAM and system RAM during inference, as it would be too slow.
For some definition of "typical", sure, but it depends on what performance trade someone is willing to accept. Strata LLM is geared around streaming weights as needed. It lets me run DeepSeek-4.1-Flash 1031, heavily quantized (ISTA-DASLab's 2.25 bpw quant), on a single Radeon R9700 plus 256 GB host RAM. 500 tps prefill and 30 tps decode isn't great, but it's a large fraction of what I get with Qwen3.8-27B quantized to fit entirely in VRAM for short prompts, it's faster past about 80k context, and DS4.1F can use a 1M-token context. It takes forever to fill that, or even to refill if the KV cache doesn't match, but it works.
Something doesn't seem right. Vanilla llama.cpp can hit 50 t/s on an R9700. Q6 (albeit with limited context of 64k, more if kv cache is quantized) fits entirely on the card.
I hear what you're saying - there are people who will use it even at 0.1 tokens/s and run tasks over night or over multiple days if it is good. but Qwen3.8 Flash Next is a MoE model (6B active) so the memory bandwidth/compute required to calculate each token is very different from a 65B (meaning, 65B all active) dense model!
Is it competitive overall or just competitive when compared to "U.S. Open-Weight models that are otherwise pretty decent at agentic coding"? The latter does not imply the former.
What'll be interesting is how it performs relative to Qwen3.8 27b when quantised to the same size. Which, of course, is a comparison they don't want to do.
Well, obviously. These newer American models aren’t competing with Qwen yet. They are competing for American businesses and organizations that can’t/won’t use Chinese models.
They are playing catch-up, but I’m happy to see new players to the game, and it will be interesting to watch how they develop
Congratulations I guess but it's a pretty bad look to not even mention those other open weight models.
What's funnier is that after cherry picking models for comparison, they further cherry picked Inking for size comparison instead of better and smaller Beam 1.0
The bar chart also shows Nemotron 3 Ultra instead of smaller and better Nemotron 3.5 Super.
We've always been at war with eurasia.
Seriously though, it's just easier to pretend that China doesn't exist when they're stomping us this badly.
I think it's because the primary users of this model won't even look at "those other open weight models" as options. This is the right chart to present to those who would consider this model. (e.g. US Gov't agencies, chip design industry, etc etc)
Right? These are not likely to be meaningful points of comparison to anyone but the authors.
Maybe Palantir will use it or something. There probably is a market for US models but I don't know how large it will be.
US Gov won't let Chinese models be hosted on any internal infra, not even supposed to use them for local AI. So there is absolutely a market for US-made or non-Chinese models there.
Because Chinese models heavily rely on distillation and regularly behave strangely in public benchmarks (where they score much better than in private evaluations), even if many world class researchers are Chinese. It's difficult to quantify to what extend labs like Moonshot would be competitive if they could not distill from OpenAI, now that Anthropic mostly blocks it
https://www.theinformation.com/articles/openais-astra-model-...
Do those other models usually mention all the competition? I haven't noticed
This model scores 80.5 on Terminal Bench 2.1
For reference Qwen 3.8 Flash Next (which is bigger than this model but probably cheaper to run due to MoE) scores 84.3, and Qwen 3.8-27B scores 73.0.
I think this model could be competitive, depending on how much it was benchmaxxed compared to Qwen?
There are several situations where some of us are explicitly prohibited from using models from companies not under the jurisdiction of the United States. It doesn't mean these models are competitive when looked at from a global view but it is also not an irrelevant differentiation.
It's the same positioning Mistral makes.
I trained a tiny LLM earlier this year, it's fits under 10mb and it's completely useless, can I call it SOTA for my neighborhood?
Speaking as someone who has High-Memory Mac Studios, it is great that there is finally a model that will run at usable speeds with high quality. In my experience, the sweet spot for Mac Clusters is around 200GB of weights, which this is far under. Congratulations to the recursive team for building an open model that will actually run!
It will be slow on Macs, it's a dense model.
When a post doesn’t mention Qwen for comparison, you know the performance isn’t going to be great.
I mean Qwen 3.8 27B hits 73% on TB2.1 and 42.2% on DeepSWE....
Agentic coding benchmarks look great.
I think we will find the market is going to move into omni multi-modal models: text, moe, agentic tasks, voice, images, and video all in one LLM.
I understand the desire to look the best in some self defined category. But not showing any comparison against the actual leaders in open weight renders all of the benchmarking completely information-less for the actual end users (who care about real performance comparison against the real field).
Global open-weight SOTA ... https://huggingface.co/moonshotai/Kimi-K3.1-Pro
These are pretty odd benchmarks to use. Terminal Bench 2.1 is old. SWE-bench was disavowed by OpenAI. DeepSWE is clearly benchmark-hacked by Gemini.
So we have a Recursive and a Ricursive, both making LLMs?
Time to start Rickursive, the lab that will never give you up and never let you down.
And it should acquire the existing two
Roll them up, you might say.
we need RichUrsive
One making self-improving AI via AI-chip co-design, the other relying only on software for self-improving AI.
I hope they never have to call each other about anything. It could get confusing if neither knows who is the base case.
There seems to be some recursion going on here.
Not comparing to Qwen3.8 27B gives cherry-picked benchmark comparison energy. Still, Ripple-65B outscores Qwen3.8 27B, according to the reported benchmarks.
Showing a terminal bench 2.1 score at the top is very misleading when terminalbench 4.0 is the current metric as a very quick glance for anything we care about for 30B to 180B size possible to locally-host models.
Ouh this is the size of model I want to see more of (just got a 98gb m5u studio)
Not to be confused with https://recursive.com/
Would love a 27B model. Wonder if we'll get a distillation.
Ouh that fits right on a base M5U Studio! Excited to try it.
The comparison chart is befuddling.
So the model is "meh" at best, but the cost of the training is interesting. So we can extrapolate that in 3 years the cost will be below $5m for a model like this.
It's very well within the capabilities of any reasonable organization, maybe even crowd-funded.
What is this thing where model comparisons blatantly pick other crap models and show themselves at the top left (cheap and effective)? Put all the open models, or put all the cheap models, or whatever, but just cherry picking is egregious.
I'm sorry but what is "U.S." implying here?
Is this sponsored by the us gov? did it only train on US data? does it only work on the US?
It's a US group that made it, in the same vein that OpenAI and Anthropic have US models.
It means that it's an American AI lab behind it.
...hiring Chinese people to do it.
Looks like all their engineering jobs are in Palo Alto?
Palo Alto is full of Chinese people.
I think it’s the same as EU sovereign ones?