News for hackers. Stuff that matters.
Re:Mercury 3: Diffusion at scale by volodia on Monday October 19, 2026 @01:37PM (#50155371)
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.

Re:Mercury 3: Diffusion at scale by ttul on Monday October 19, 2026 @01:53PM (#50155574)
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.
Re:Mercury 3: Diffusion at scale by volodia on Monday October 19, 2026 @03:02PM (#50156496)
We might not be able to precisely explain the 6% difference on that specific benchmark, but overall we have found the model to be Haiku- or Luna-level in our own daily usage. That means that it won't yet replace Opus or Fable as the primary driver of a harness, but we have used it successfully for subagents, including implementing kernel optimization ideas produced by a larger model (and those improvements have gone into production -- a form of RSI).

In general, the model continues to improve when given additional high quality data like a normal LLM, which is what makes us very bullish on the scaling of dLLMs.

Re:Mercury 3: Diffusion at scale by ttul on Tuesday October 20, 2026 @09:55AM (#50166531)
Super cool. I'll have to give it a whirl! Keep up the amazing work.
Re:Mercury 3: Diffusion at scale by LarsDu88 on Monday October 19, 2026 @02:26PM (#50156011)
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?
Re:Mercury 3: Diffusion at scale by volodia on Monday October 19, 2026 @02:56PM (#50156414)
We've been happy with how performance has scaled with batch size. The model runs in our API at reasonable costs, which is also what allows us to serve the model with competitive prices ($0.10 per 1M input and $0.35 per 1M output at launch).
Re:Mercury 3: Diffusion at scale by LarsDu88 on Monday October 19, 2026 @03:57PM (#50157260)
Pricing can always be set at a loss, however, particularly for a private company. Are there solid published numbers for a specific hardware spec?
Re:Mercury 3: Diffusion at scale by AmazingTurtle on Monday October 19, 2026 @01:57PM (#50155617)
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?
Re:Mercury 3: Diffusion at scale by volodia on Monday October 19, 2026 @02:51PM (#50156345)
We're not releasing the weights, but the model is available using a standard API, and it integrates with your favorite harnesses.

Here is a page that compiles the list of integrations: https://docs.inceptionlabs.ai/integrations/overview

For example, it's just a few small steps to run the model in Cursor: https://docs.inceptionlabs.ai/integrations/cursor

You shouldn't expect Opus-level performance yet, but early testing has found the model to be great for coding subagents.

Re:Mercury 3: Diffusion at scale by jacquesm on Monday October 19, 2026 @01:53PM (#50155573)
Will you make the weights available?
Re:Mercury 3: Diffusion at scale by volodia on Monday October 19, 2026 @03:05PM (#50156549)
Not at the moment, but we are looking to share more information about the model in the future via technical blog posts and technical reports.
Re:Mercury 3: Diffusion at scale by jacquesm on Monday October 19, 2026 @03:05PM (#50156562)
Ok. Good luck then.
Re:Mercury 3: Diffusion at scale by himata4113 on Monday October 19, 2026 @01:49PM (#50155512)
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!
Re:Mercury 3: Diffusion at scale by volodia on Monday October 19, 2026 @03:08PM (#50156594)
We are going to be sharing additional information like task completion times in the near future as the model undergoes additional benchmarking by third parties. We also plan to share data on specific use cases where the model traces new points of the Pareto frontier and and speed cost.

For a preview of the latter, check out our analysis of the Mercury Voice model: https://www.inceptionlabs.ai/blog/introducing-mercury-voice

Re:Mercury 3: Diffusion at scale by Topfi on Monday October 19, 2026 @02:58PM (#50156444)
If I may, are there plans for visual input support?
Re:Mercury 3: Diffusion at scale by volodia on Monday October 19, 2026 @03:03PM (#50156518)
We have internal prototypes, so please reach out to [email protected] if you're interested in early access
Re:Mercury 3: Diffusion at scale by 0xb0565e486 on Monday October 19, 2026 @01:56PM (#50155604)
Incredible. Any luck in having a model that's multimodal in the future?
Re:Mercury 3: Diffusion at scale by volodia on Monday October 19, 2026 @03:04PM (#50156535)
This is an area that we're really excited about. We have internal prototypes, and if you're interested in trying it out as an early tester, please reach out to [email protected]
Re:Mercury 3: Diffusion at scale by WithinReason on Monday October 19, 2026 @03:01PM (#50156493)
Is the model size secret?
Re:Mercury 3: Diffusion at scale by volodia on Monday October 19, 2026 @03:26PM (#50156847)
We're not publicly announcing the model size right now, but we can compare against other models with a similar intelligence -- what we see is a big boost in speed for the same level of capabilities.
Re:Mercury 3: Diffusion at scale by lostmsu on Monday October 19, 2026 @01:54PM (#50155588)
I don't see it on Baseten. Is it special in some way?
Re:Mercury 3: Diffusion at scale by krooost on Monday October 19, 2026 @02:10PM (#50155808)
Re:Mercury 3: Diffusion at scale by lostmsu on Monday October 19, 2026 @02:17PM (#50155898)
Not listed among https://app.baseten.co/model-apis

Does it mean dedicated inference only?

Re:Mercury 3: Diffusion at scale by krooost on Monday October 19, 2026 @02:54PM (#50156382)
You have to reach out to Baseten directly for access. You can also access through our API (https://platform.inceptionlabs.ai/) or OpenRouter.
Re:Mercury 3: Diffusion at scale by johanyc on Tuesday October 20, 2026 @03:31AM (#50162532)
https://openrouter.ai/inception/mercury-3#performance

It seems the latency is not very good. Is this a temporary compute problem or is it inherent to diffusion?