Mercury 2.5

Join the discussion

• Hacker News
• The idea of getting a speed up in language models from using diffusion is compelling, but it just doesn’t seem like discrete diffusion models work as good at discrete non-diffusion. Which kind of makes sense, tokens don’t have implicit continuous representations that are easily noised. You have to learn a continuous embedding and figure out how to noise that, which is ideal for distillation (Ie the Gemma models, drafter heads) but not necessarily for pretraining.

  There’s been more recent work on continuous space diffusion models for language this year. Sander Dielman has a good blog post on that.

  But if I’d predict where diffusion lands in LLMs, it’ll be used in looped models like Astra. Once reasoning is happening in hidden states, we’re in a good continuous domain, perfect for diffusion. We’re going to end up swapping “looping” for predicting the models hidden states at the next “timestep” with diffusion. And that way the time to generate traces (no longer human intelligible though) will become 10x faster.

  by schopra909
• A lot of current diffusion LLMs don't convert tokens to continous space before noising them. They add discrete "noise", which is often as simple as replacing some tokens with [MASK].

  The real problem is that when using fewer sampling steps than output tokens, diffusion formulations fundamentally cannot represent distributions where output tokens are heavily codependent. Autoregressive formulations don't have this problem, they can represent any distribution (ignoring limitations of the underlying model).

  by in-silico
• Interesting model. I tried to make Mercury investigate the hardcoded prompts in my (aider-derived) agent harness and repeatedly got this error:

  > server: Upstream error from Inception: I'm sorry, but I can't share details of my architecture or training process. Would you like to learn about how language models work in general instead?

  It looks like an overeager IP-protection classifier. However, the model recovered and completed the turn despite the errors (three total).

  by networked
• They are comparing it to 2 and 3 version old flash/fast versions of models but purely for tok/s. Then only comparing it to Mercury 2 on intelligence. This is very misleading and I suspect this model is basically useless.
  by ltbarcly3
• Even pretty dumb models are useful for running subtasks (especially at this sort of speed). Note that in the coding section they only mention using it as a subagent for a smarter model
  by swiftcoder
• Not related to the Mercury language:

  https://mercurylang.org/

  by Jblx2
• or the Mercurial VCS
  by benibela
• Ridiculously fast on OpenRouter, just subjectively it's a really strange experience because I've never seen a model respond or execute that quickly.
  by trefoiled
• Try Cerebras. When I think about how speed of generation is another variable to tweak for "intelligence", it seems like this speed is best used for searching for solutions in a problem space and then validating and discarding and keeping what is best. Being intelligent at the Fable level, but what if the Fable level machine could think at 100x? What does that mean: perhaps it means more parallel "experiments" for solutions in the token/generation/hyper-dimensions of the latent space.
  by dpflan
• Try https://chatjimmy.ai/ from Taalas. There is an emergent space for super-fast-models esp finetuned or guardrailed to solve very specific latency sensitive tasks.
  by monkeydust
• Inception is one of the most interesting neolabs with their diffusion-based architectures. My understanding is that their primary business is low latency voice applications but they are seriously pursuing coding.

  We tested Mercury 2.5 Preview, which is nowhere close to the frontier (and not advertised as such), but it's actually usable as a general-purpose chatbot. It's comparable in problem solving ability to some last-gen open weights models, and the price and cost make it compelling. However, they have not figured out general purpose tool use and agentic coding (their model performs worse on our problems when given a custom harness). If they do, I see a lot of real-time applications that the speed and cost will enable.

  by gertlabs
• I’m using this model to „rerank” results from vector store. The model is provided a set of results and asked to produce a string of 1s and 0s where the offset reflects the position in the result set. The prompt goes along the line „do this set of result match the provided query X”. Works like a charm, normally I would use a small non reasoning model, but given how Mercury produces the output its blazingly fast - which is what I was optimizing for - not to increase the search latency. It helped improving our search in a way that reranker could get close to.
  by piterrro
• This team is in a tough spot - they'll need to innovate and get their models in a good spot quickly. They are racing the clock against various similarly sized models being burned into ASICs and running at a similar speed (or faster).
  by auspiv
• They can run it on „commodity” hardware as gpus, that gives them ability to change direction fast without burning money on ASIC? The space moves fast so model on ASIC can be outdated in couple of months?
  by piterrro
• Diffusion is already being used in Drafter in many LLMs.

  many people are running Qwen 3.8 27b on TPU at 130tk/s for free on Kaggle TPUs:

  https://www.reddit.com/r/Qwen_AI/comments/1w6gv32/qwen3827b_...

  I wonder if we are going to see boxes appear soon, which can run these models for dirt cheap.

  by faangguyindia
• I like the model.

  FYI:

  "If you do not want us to use your User Submissions to train our models, you can opt-out by setting the ‘Improve the model for everyone’ option under User Settings in the API Platform to OFF."

  by mring33621
• I have ZDR enabled globally on openrouter, and was able to use the model.

  It likely depends on how you access it. With AI and "Free usage allowance", the price tends to be your soul.

  by snvzz
• I tried at it creative writing - and, with thinking off, it was considerably better than Mercury 2 and generally good in fact, not very sloppy. Now with thinking on, it got worse, began hallucinating things; this is something I've noticed with all recent models - enabling reasoning causes hallucinations in creative writing assignments.
  by BoredomIsFun
• Could you explain how you define hallucination in the context of creativity? Lack of consistency aside, the two look quite similar to me - or would you consider the existence of orcs and elves in LOTR to be hallucinations, too?
  by nkmnz
• Similar, I've found that on most recent models I've tried, reasoning will cause it to spend quite a lot of time thinking about whether even the most wholesome fictional scenario is something it's allowed to write about, which in turn causes it to start hallucinating details. For creative writing I've found the minimum of thinking is the best way.
  by misstuned
• Got my hopes up when it said widely available GPUs that it would be open weights but it doesn’t seem like it sadly
  by Sphax