I created a pelican worked wonderfully, better than old haiku model or even sonnet models, maybe slightly worse than new gen models.
But overall it's a pretty ok model, the issue with speed seems to be that ttft is still high but still not sure how high thinking burden would be, I hope it scales well will be using it a lot if it works well.
This might actually crush most other LLMs if it's price and performance actually scales well. It was as cheap as deepseek for a few tasks I tried.
Anyone know if there is a subscription for it? I might want to test it out properly without burning a few hundred dollars on it.
Their preprint showed it worked well with small batch size, but most prod deployments right now have rather large batch sizes from serving multiple users
Honestly, a lot of those issues could be remedied with more optimized kernels, which LLMs could definitely write quite well now. I wouldn't be surprised if some deployments of Gemini Flash switched to diffusion for cases where there is low utilization, since Google showed you don't need to retrain an autoregressive model to rework it for diffusion inference.
This looks noticeably more capable than previous gen. I purposely gave it a bit of a vague prompt to create an html renderer for UML diagrams, it produced impressive work. Then i asked a question, and in the third prompt I asked it to update the html which it did correctly.
Another thing that stood out, and this might just be a novelty thing, is that it picked different and less painful-to-eyes color schemes for each theme. Congrats to the team behind it.
Question (since their team has reps here and I have never used Inception via api): when you ask for a code change in your codebase (via coding agent), how much code does it need to read? How well does it handle tool calls? Since diffusion is a different way of doing things, I am curious about the code-needed for each incremental change in the codebase, say over a larger task.
I'm very curious what the actual model size is and how the training process differed from a traditional LLM. Cool work regardless, incredible how catching up to a few months behind the cutting edge now seems like an achievable feat for new entrants, if still quite costly.
I'd be surprised if major labs aren't doing any diffusion experiments, and if they are, it would also be surprising if another lab is ahead of them in diffusion given just how much more compute they have at their disposal. My intuition could be way off though.
Google more or less ordinated the category. They have now seemingly dropped it, ostensibly because it doesn’t scale economically like transformers do (via batching requests).
I think though, that if Inception can keep up these speeds and prices, assuming diffusion continues to improve (likely), that Flash and Luna could lose out for many use cases.
If their benchmarks are to be believed they've made some real progress. I hope we see more open weight diffusion LLMs (currently there's diffusion gemma). Also, if diffusion LLMs are starting to come into their own that's going to be somewhat disruptive isn't it? If these are running at 10x throughput (at least) then that's a lot more tokens on the same hardware.
EDIT: just trying this out on a repo and it's actually pretty good so far. Pleasantly surprised.
I wonder given its speed how competitive it would be with the current batch of decision models like Jev et al. Seems to me for decision models you really want maximum throughput.
Are these models harder to reason about given they produce tokens non-linearly? Especially since we rely on non-latent reasoning to know what the model is "thinking".
"Quality: On par or better than frontier speed-optimized models like GPT-6 Luna and Claude Haiku 5.5.
Speed: 1,068 tokens per second on NVIDIA GPUs.
Context: 1M tokens, up from 260K for Mercury 2.5
Price: $0.20 per million input and $0.70 per million output tokens.
At launch, Mercury 3 is 50% off at $0.10 per 1M input and $0.35 per 1M output. Cached tokens are 90% off the input price."
Hi HN, this is Volodymyr, Inception Co-Founder & Chief Scientist.
Today we're launching Mercury 3, our most capable production diffusion LLM.
Quality: Outperforms GPT-6 Luna and Claude Haiku 5.5 across a range of benchmarks at matched reasoning effort.
Speed: 1,000+ tokens per second on widely available NVIDIA Blackwell GPUs, 4–8x faster than those models.
Context: 1M tokens, up from 260K.
Price: $0.10 per million input and $0.35 per million output tokens at launch.
The main takeaway is that diffusion scales. Since Mercury 1 we’ve scaled the model more than 40x in total parameters, and intelligence has gone up at every step without giving up speed.
Diffusion generates tokens in parallel through iterative denoising rather than one at a time, so as reasoning models think longer per answer, the speed advantage over autoregressive models of the same size grows.
Early-access partners are running Mercury 3 in browser agents, tutoring apps, real-time search, and customer support. It’s OpenAI API compatible, and it’s also available on OpenRouter, Baseten, and Vercel.
Try it on our API with 100M free tokens.
Happy to dig into the architecture, scaling results, evals, or anything else in the comments.
Google's research showed some of the speed benefits tended to drop off at larger batch sizes on an H100. Are there published batch-size vs tokens/s figures for Mercury 3?
I'm curious if you could comment on why your DeepSWE performance is below Haiku. This is not a dig - getting 46% vs. Haiku's 52% is still amazing. I'm just wondering if perhaps your post-training was less aggressively RL'ed on coding tasks than perhaps Haiku was. Having a cheap 1,000 tok/s option for coding work is a dream because - if it's reliable enough - you could blast through work such as high-level security reviews or finicky and repetitive refactorings that don't necessarily require SOTA performance.
Whats the model total param size and will you release the weights?
What does harness integration look like? Do you have an harness? Is it good at coding? What about tool calls? Do we have to wait a fixed amount of time for every step it takes?
Why medium in benchmarks? I've not seen that many people use other than xhigh/max on the cheap models. No low->max performance curve either. Time to task completion would also be nice!
I created a pelican worked wonderfully, better than old haiku model or even sonnet models, maybe slightly worse than new gen models.
But overall it's a pretty ok model, the issue with speed seems to be that ttft is still high but still not sure how high thinking burden would be, I hope it scales well will be using it a lot if it works well.
This might actually crush most other LLMs if it's price and performance actually scales well. It was as cheap as deepseek for a few tasks I tried.
Anyone know if there is a subscription for it? I might want to test it out properly without burning a few hundred dollars on it.
Pelicans aren't bad, a little quirky: https://tools.simonwillison.net/markdown-svg-renderer?url=ht...
Those were via OpenRouter - here's another set I made by calling the Inception API directly: https://tools.simonwillison.net/markdown-svg-renderer?url=ht...
I still don't understand why Google did effectively nothing with their Gemini Diffusion demo from last year.
It was really cool, and really fast, and just went quiet.
They did release a Diffusion Gemma a few months ago but I've not seen much of a splash from that one, either.
Their preprint showed it worked well with small batch size, but most prod deployments right now have rather large batch sizes from serving multiple users
That's a great answer, thanks. I'd missed that detail.
Honestly, a lot of those issues could be remedied with more optimized kernels, which LLMs could definitely write quite well now. I wouldn't be surprised if some deployments of Gemini Flash switched to diffusion for cases where there is low utilization, since Google showed you don't need to retrain an autoregressive model to rework it for diffusion inference.
They probably did not get it to work well?
Good at research, terrible at product?
This looks noticeably more capable than previous gen. I purposely gave it a bit of a vague prompt to create an html renderer for UML diagrams, it produced impressive work. Then i asked a question, and in the third prompt I asked it to update the html which it did correctly.
Another thing that stood out, and this might just be a novelty thing, is that it picked different and less painful-to-eyes color schemes for each theme. Congrats to the team behind it.
Question (since their team has reps here and I have never used Inception via api): when you ask for a code change in your codebase (via coding agent), how much code does it need to read? How well does it handle tool calls? Since diffusion is a different way of doing things, I am curious about the code-needed for each incremental change in the codebase, say over a larger task.
I'm very curious what the actual model size is and how the training process differed from a traditional LLM. Cool work regardless, incredible how catching up to a few months behind the cutting edge now seems like an achievable feat for new entrants, if still quite costly.
I'd be surprised if major labs aren't doing any diffusion experiments, and if they are, it would also be surprising if another lab is ahead of them in diffusion given just how much more compute they have at their disposal. My intuition could be way off though.
Google more or less ordinated the category. They have now seemingly dropped it, ostensibly because it doesn’t scale economically like transformers do (via batching requests).
I think though, that if Inception can keep up these speeds and prices, assuming diffusion continues to improve (likely), that Flash and Luna could lose out for many use cases.
@volodia (Inception co-founder) has a comment here which is flagged/dead. That's weird; it shouldn't be. Can someone fix that?
Not as fast as GPT 6.1-Sol on my favorite toy math puzzle. It might be one of those things where intelligence consumes less tokens
Hmm at least in Openrouter, deepseek v4.1 flash via together/baseten/fireworks still beat mercury 3 on latency and intelligence
Ttft is quite slow with mercury
If their benchmarks are to be believed they've made some real progress. I hope we see more open weight diffusion LLMs (currently there's diffusion gemma). Also, if diffusion LLMs are starting to come into their own that's going to be somewhat disruptive isn't it? If these are running at 10x throughput (at least) then that's a lot more tokens on the same hardware.
EDIT: just trying this out on a repo and it's actually pretty good so far. Pleasantly surprised.
I wonder given its speed how competitive it would be with the current batch of decision models like Jev et al. Seems to me for decision models you really want maximum throughput.
They have a separate decision model called Mercury Decide
Looks excellent, looking forward to trying this out. Feels like we're reaching a critical point where these become competitive.
Damnit, I've been waiting on an open weights model like this one, still no dice :(
diffusion gemma?
Was mostly an experiment. You're not going to use it for any real coding. Hopefully we'll see another updated diffusion gemma in the future.
Are these models harder to reason about given they produce tokens non-linearly? Especially since we rely on non-latent reasoning to know what the model is "thinking".
I really wish they would make these multimodal. Seems like a definite improvement over 2.5 though in my testing
100M free tokens? wow.
I wish they had a coding plan so your could kick the tires before paying API rates.
"Quality: On par or better than frontier speed-optimized models like GPT-6 Luna and Claude Haiku 5.5. Speed: 1,068 tokens per second on NVIDIA GPUs. Context: 1M tokens, up from 260K for Mercury 2.5 Price: $0.20 per million input and $0.70 per million output tokens. At launch, Mercury 3 is 50% off at $0.10 per 1M input and $0.35 per 1M output. Cached tokens are 90% off the input price."
Hi HN, this is Volodymyr, Inception Co-Founder & Chief Scientist.
Today we're launching Mercury 3, our most capable production diffusion LLM.
Quality: Outperforms GPT-6 Luna and Claude Haiku 5.5 across a range of benchmarks at matched reasoning effort. Speed: 1,000+ tokens per second on widely available NVIDIA Blackwell GPUs, 4–8x faster than those models. Context: 1M tokens, up from 260K. Price: $0.10 per million input and $0.35 per million output tokens at launch.
The main takeaway is that diffusion scales. Since Mercury 1 we’ve scaled the model more than 40x in total parameters, and intelligence has gone up at every step without giving up speed. Diffusion generates tokens in parallel through iterative denoising rather than one at a time, so as reasoning models think longer per answer, the speed advantage over autoregressive models of the same size grows.
Early-access partners are running Mercury 3 in browser agents, tutoring apps, real-time search, and customer support. It’s OpenAI API compatible, and it’s also available on OpenRouter, Baseten, and Vercel.
Try it on our API with 100M free tokens.
Happy to dig into the architecture, scaling results, evals, or anything else in the comments.
Google's research showed some of the speed benefits tended to drop off at larger batch sizes on an H100. Are there published batch-size vs tokens/s figures for Mercury 3?
I'm curious if you could comment on why your DeepSWE performance is below Haiku. This is not a dig - getting 46% vs. Haiku's 52% is still amazing. I'm just wondering if perhaps your post-training was less aggressively RL'ed on coding tasks than perhaps Haiku was. Having a cheap 1,000 tok/s option for coding work is a dream because - if it's reliable enough - you could blast through work such as high-level security reviews or finicky and repetitive refactorings that don't necessarily require SOTA performance.
Whats the model total param size and will you release the weights? What does harness integration look like? Do you have an harness? Is it good at coding? What about tool calls? Do we have to wait a fixed amount of time for every step it takes?
Why medium in benchmarks? I've not seen that many people use other than xhigh/max on the cheap models. No low->max performance curve either. Time to task completion would also be nice!
Will you make the weights available?
Incredible. Any luck in having a model that's multimodal in the future?
I don't see it on Baseten. Is it special in some way?
https://www.baseten.co/library/mercury-3/
Not listed among https://app.baseten.co/model-apis
Does it mean dedicated inference only?