Gemma 4 12B: A unified, encoder-free multimodal model

(blog.google)

220 points | by rvz 2 hours ago ago

85 comments

  • minimaxir an hour ago

    The big story here is the encoder-free part, which I still don't fully understand.

    > Vision: We replaced Gemma 4’s vision encoder with a lightweight embedding module consisting of a single matrix multiplication, positional embedding and normalizations.

    That's technically encoding, just without using a dedicated model for it like SigLIP? The Developer's Guide elaborates, it's still a 35M layer which I am curious is robust enough. https://developers.googleblog.com/gemma-4-12b-the-developer-...

    > Small enough to run locally on consumer laptops with 16GB of RAM, it unlocks powerful multimodal and agentic experiences right on your machine.

    I am assuming that involves quantization, which due to the quality loss makes that statement somewhat misleading IMO.

    • spott 11 minutes ago

      This is just early fusion basically.

      FAIR did this 2 years ago now: https://arxiv.org/abs/2405.09818

      I've been waiting for something like this to be released since then.

      The annoying thing is that chameleon was multi-modal out based on the same principles, but this model is just inputs... (I'm curious how they did pre-training without having multi-modal outputs as well. I wonder if they just chopped them off rather than support image output).

    • georgehm 33 minutes ago

      Embedded within that developer page is a good explainer of the encoder free architecture . https://newsletter.maartengrootendorst.com/p/a-visual-guide-...

    • mchinen 19 minutes ago

      The audio side is even more interesting, as it seems they totally got rid of positional embedding are just doing a single linear transform to match the LLM input dimension and that's it.

      > Audio: We simplified audio processing even further. We removed the audio encoder entirely and projected the raw audio signal into the same dimensional space as text tokens.

      • make3 6 minutes ago

        I guarantee you there's positional information one way or another. they just don't mention it because positional embeddings are extremely cheap computationally, not worth mentioning

    • woadwarrior01 5 minutes ago

      There are many priors to encoder-free VLMs. I specifically remember the EVE series of models from ~2 years.

      https://github.com/baaivision/EVE

    • jszymborski an hour ago

      Totally agree that it is "encoding" in the general sense, but I think they are referring to the lack of an "encoder" neural network.

      • minimaxir an hour ago

        In hindsight I may have been pedantic.

        • wilkystyle 40 minutes ago

          I had a similar thought to you, and found your question and the resulting discussion helpful!

        • alberto467 36 minutes ago

          Not at all. Getting really pedantic, tokenization is also a form of encoding, so it doesn't matter the modality you're using, you'll end up doing some type of encoding in some way.

    • kristjansson an hour ago

      > quantization

      12b means 12G @ 8 bits/param (basically lossless) and 6G at 4 b/p (generally accepted 'pretty close' level). Not too bad?

      But TBD how well the base model performs before thinking too much about quantization

    • matja 41 minutes ago

      One side-effect, is that the separate .mmproj file (Multi-Modal Projection encoder) is no longer needed, when using the model with llama.cpp etc.

      • pferdone 10 minutes ago

        But do I have the option to run it 'text only'?

    • rao-v 19 minutes ago

      Encoder free is huge for running on SBCs etc. often the encoding time is a significant fraction of generation time if you are using a VLM as a all purpose vision model

    • reactordev an hour ago

      It actually works well because unlike encoders, the latent space is trained on that initial layer so it ā€œknowsā€ what to do with that sparse density. I’ve been using gemma4-12b with Flux2 and its ability to reason on visual input is pretty good. That said, each model is good in their own ways so YMMV but overall, it’s about as solid as Qwen just with a more advanced architecture.

    • wolttam an hour ago

      I think the idea is that the model is seeing embeddings that map directly to underlying pixel data, rather than being fed semantically rich embeddings from an encoder model which itself had seen the raw pixel data.

    • LarsDu88 an hour ago

      Well its a real simple encoder I guess

    • GaggiX an hour ago

      > That's technically encoding

      Isn't that just projecting the patches into the d_model size vectors that the models takes?

      >I am assuming that involves of quantization

      12B model in 16GB seems very reasonable to me, int8 is top quality for running models.

      • minimaxir an hour ago

        The guide describes it as projection although there is apparently an extra step: "A factorized coordinate lookup (X and Y matrices) attaches spatial location information directly to the input."

        12B at int8 would take up 12G memory, or 75% of the system memory which technically fits within 16GB but the OS will not like that.

  • ethanpil an hour ago

    What's Google's business case for releasing open models? Don't get me wrong, I am grateful and appreciative of these releases. I'm trying to understand how it fits into their bigger picture as a for profit company? Are they not helping competitors build on the novel technology they have developed?

    Is it simply goodwill and/or marketing? Or am I missing something strategic?

    • browningstreet an hour ago

      This won't replace commercially viable, revenue generating alternatives of their own devising, but it does enable development activity and initiate conversations with enterprises who start with this model but want to do slightly more.

      That's my experience right now... my company is all in on a plethora of platform products. Also, Microsoft just yesterday said their goal was "Unmetered intelligence". There's a lot of things that can be enabled by small local models, and those things are part of stacks that can generate revenue in other layers.

    • gen220 24 minutes ago

      A big part of the frontier labs abilities to charge 80% gross margins on inference is having the cornered resource of frontier models.

      If that inference becomes popular and valuable enough that those companies make billions of dollars in profit, those companies could use that profit to fund the building of alternative products and platforms that dis-intermediate google's relationship with the customer.

      Google already has an 80% gross margin business, the biggest one in the world. Everybody wants a slice of it.

      By offering frontier inference closer to cost and open-sourcing everything that's sub-frontier, they're commoditizing frontier labs' models, which inhibits their ability to durably make high gross margins on inference.

      It's a strategic play.

      • zozbot234 21 minutes ago

        A 12B-sized model is a far cry from "frontier inference". That's more like DeepSeek V4 Pro territory which is a 1.6T model. Or for multi-modal models, Kimi 2.6 which is 1T.

        • gen220 15 minutes ago

          at risk of quoting myself... :)

          > By offering frontier inference closer to cost *and* open-sourcing everything that's sub-frontier

          It's two prongs! One prong is that their frontier inference pricing is significantly cheaper/closer-to-at-cost as Anthropic's.

          The subject of this thread is the other prong: offering compelling models that are sub-frontier and self-hostable.

          Self-hosting models and at-cost frontier models are the high-end and low-end disruptions, respectively, to Ant/OAI/etc.'s business models.

          • echelon 2 minutes ago

            Google needs an anti-trust breakup about 10 years ago.

            They need one more than ever now.

            This is ridiculously anti-competitive.

        • boutell 14 minutes ago

          You're right that it's not literally frontier. But like recent Qwen releases, it is a lot more capable than anybody thought models of this size could be a year ago, like capable enough to set a ceiling on what you can charge for AI for certain applications. Others still clearly justify a stronger model, but this trend may continue, etc.

    • Mr_P an hour ago

      Android and Chrome need on-device AI capabilities. Google can't lock down those weights like it can with server-side ML.

      So it's easier to just release those models as open source and make it official, since someone would inevitably hack the weights out anyway.

      • Aachen 40 minutes ago

        Could say the same for camera processing in the Pixel Camera app or any other binary someone wants to re-use that comes included in a software distribution (seemingly for 'free'). They can't lock the instructions up on the server so they might as well make the binary be freely distributable?

        Companies don't commonly give away executable binaries "just because", why'd they start now for these binary blobs that are the models?

        Not that I'm unhappy about it! Yay for open data any day, I'm just not understanding why, at least beyond PR in nerd circles

        • jack_pp 21 minutes ago

          Because a model like this can't be as easily obfuscated as image processing. Image processing is a bundle of many moving parts, a lot of functions each with it's own inputs and outputs. A model is a single function which can be easily extracted and reused, in comparison

    • beambot 40 minutes ago

      Google is one of the few verticalized options in AI: Data, models, cloud services, low-level silicon (TPUs), internal use cases, retail use cases, B2B uses, distribution (browser & mobile), etc.

      They rise with the tide of AI adoption. But they gain ground if people opt into Google solutions. And any token sent to a Google model (free or paid) actively punishes their competitors that are then required to spend vast sums to remain bleeding edge.

    • onlyrealcuzzo an hour ago

      If you're an AI lab, you definitely want research teams in this space - as this is where you can most easily iterate and make improvements which you'll then bake into larger, frontier models.

      The question is: do you want to release your models, or use them purely for R&D?

      Since everyone else is already releasing models of similar qualities, it's hard to say you're shooting yourself in the foot if you join the chorus.

      The added cannibalization of releasing them is effectively zero, so the reputational benefits are likely to be worth it.

    • staticman2 24 minutes ago

      As long as Chinese firms are releasing good open models I imagine there isn't a huge downside for Google to release state of the art small models to compete in the "free" space.

    • estearum an hour ago

      It's to destroy possible footholds for competitors and prevent them from making money in segments that Google doesn't care too much about, but can trivially commoditize.

    • rootusrootus an hour ago

      Neutering OpenAI and Anthropic would be my guess. Commoditized LLMs won't hurt Google nearly as much as it hurts the LLM-only companies, and so accelerating the inevitable just helps knock out potential future competition in areas where Google -does- make a lot of money now.

      • literalAardvark 15 minutes ago

        I think this plays a part, but the truth is that Google doesn't need to do that, Chinese open models are already doing that by themselves.

        So perhaps another part is just Google showing that they can indeed play at the big boys table.

    • theturtletalks an hour ago

      Maybe they are hedging against a future where local models are just as good as cloud models? Or maybe they can go the Taalas route and start hardcoding Gemma on a chip and hardware manufacturers can use it for local private AI.

    • ppeetteerr an hour ago

      Isn't Apple about to license some variation of this from google for on-device AI? Maybe it’s their sales pitch to Apple and then they will lock it down.

    • CuriouslyC an hour ago

      They're trying to capture the segment of the market that wants to control the model, with the intent of getting you to run them on Vertex.

    • stevenhubertron 43 minutes ago

      My guess is testing for Apple’s Siri replacement and partnership but that’s a total SWAG

    • mmarian an hour ago

      Marketing + Pro Serv if I had to take a guess.

    • dist-epoch 31 minutes ago

      Evangelism for AI. Google is one of the big AI providers.

      Eventually the local model is not enough, and you'll upgrade to the big ones.

    • accountrequired an hour ago

      edge compute

    • superchicken099 an hour ago

      Gemma overtakes and kills real open-source AI projects, pushing people who would support them towards enterprises like Google

    • XzAeRosho an hour ago

      Google's MO since always has been to release great products or services for free, position themselves high and then abandon them or just find uses for Enterprise sales.

      I'm pretty sure they are doing it because they get some research experience by shrinking and improving these models, and because they know that by doing this they get some good PR among the dev community.

      • Aachen 36 minutes ago

        Google's "free" is and was ad-supported, even if some products now have a paid tier. These models don't include ads. Doesn't seem like the same underlying reason

  • spott 5 minutes ago

    Is there a paper on this?

    I'm curious how they pre-trained it... I feel like it must have had audio/image output that they chopped off.

    I wonder how hard it would be to add it back on.

  • mlmonkey 6 minutes ago

    Is there some place where we can try it before downloading the gigabytes of weights?

  • ComputerGuru 31 minutes ago

    Quite aside from the architectural changes, I suppose this is the answer to why Google had such a glaring hole in the (pretrained) Gemma4 model lineup between the Gemma4 4b and Gemma4 26b models!

    A model that comfortably fits in 16GB of VRAM (allowing room for context) is a welcome upgrade.

  • lxgr 29 minutes ago

    Am I missing something or are the Ollama versions of this (https://ollama.com/library/gemma4/tags) text-only for now?

    • philipkglass 27 minutes ago

      Since ollama has diverged from llama.cpp, it will take a bit of time for ollama to support multi-modality. If you're using plain llama.cpp it looks like a PR has already merged for this model with vision and audio support:

      https://github.com/ggml-org/llama.cpp/pull/24077

      • zozbot234 7 minutes ago

        They've actually gone back to (a lightly patched) llama.cpp with the 0.30 release a few weeks ago, and have now vendored in an up-to-date release. Needless to say this is great news for both projects!

  • Havoc 35 minutes ago

    Quite a niche release. The MoE outperforms it on score and will likely be faster thanks to lower active weights. So this really only makes sense for specific ram constrained applications that can’t fit a quantized MoE

    • dist-epoch 29 minutes ago

      The un-quantized MoE outperforms it.

      But between same (V)RAM requirement 4 bit 26B-A3B and 8 bit 12B it's unclear which one will win, especially given one is MoE and the other dense.

      All the launch benchmarks are at 16 bit.

  • Zambyte an hour ago

    Is this Mac only? Or is that an Ollama issue that it only supports this release of models on Mac? It seems like every tag with the MLX badge is only supported on Mac[0], and that includes all of the tags in this release.

    [0] https://ollama.com/library/gemma4/tags

    Edit: MLX being Mac-only is independent of the model being MLX (and therefore Mac) only. The latter is what I am asking about.

    • embedding-shape 44 minutes ago

      MLX is quite literally macOS-specific technology, for other platforms you want non-MLX.

      I was sure "MLX" stood for "Metal-something-something" but can't find any reference to that somehow, anywho, "Metal" is hardware-accelerated graphics on Apple platforms FWIW.

      Edit: about the actual release on Ollama, if you're on non-Apple hardware you probably want the NVFP4 variant ("gemma4:12b-nvfp4") which was uploaded 45 minutes ago, especially if you're with a recent nvidia GPU.

    • jw1224 41 minutes ago

      MLX is Apple’s own machine learning framework, designed for Apple Silicon: https://opensource.apple.com/projects/mlx/

    • jasonjmcghee 21 minutes ago

      There's a CUDA backend for MLX now. Not sure about the maturity.

  • dwa3592 an hour ago

    This is a pretty good update. The demo video is a bit funny though - the tester asks to turn the release into bullet points. okay, the model obliges. then the tester says draft an email with this content. BAM! the LLM turns the content from bullets to passages even though it was not asked and it undid the last good thing that it did. i am not sure if it's an email etiquette to not put bullets in the email.

  • randomNumber7 an hour ago

    > Novel unified architecture: No multimodal encoders. The vision and audio inputs flow directly into the LLM backbone.

    I would be interested in how this actually works. I couldn't find a description of the model architecture (and I did check the links in the Google blog)

  • djyde an hour ago

    What are the use cases for these small models? Is there anyone using models of this scale in their daily life who could share their experience?

    • properbrew 8 minutes ago

      I think small models have a very good niche for specific tasks. I utilise a fine tuned Phi-4 model (smaller than this one) that fits in about 3.5gb of RAM (not vram) for the document processing side of things for the desktop app I develop (a bit of a shameless plug - whistle-enterprise.com).

      If you have a very specific idea for local model use you can find a way to make it work very well, you don't even need to have a graphics card or NPU chip. You just have to be extremely constrained in how it's used. I think as a generic chatbot they're not great, I'd use a hosted SOTA model and I'm a big fan of local LLMs myself.

    • robgough 8 minutes ago

      I've got a home-built dictation app that uses a local model to clear up the text and fix grammar. It was super easy to build. I’m extending it to capture meeting notes and summarise too. All on-device.

      I saw a little app the other day, I think someone posted on here, that looks at your screenshot and renames the file based off the contents of the file.

      There's tons of little examples like that. For a lot of use cases, you really don't need the frontier models.

    • Aachen 30 minutes ago

      "Small" models are the ones I can run myself on my own terms. LLMs aren't useful enough for me to justify spending hundreds of euros on a GPU with 16GB VRAM or something, and that's assuming I have the rest of the desktop just laying around. Back when I checked (before the RAM price hike), these models weren't meaningfully better than 4-8GB ones anyway, you'd have to go for the top tier cards at 24 or 32 GB iirc to get something vaguely in the direction of the SaaS versions, and that was absolutely out of my budget. Even if that changed, so have hardware prices so it'd probably still work out the same

    • Xiol 31 minutes ago

      I've yet to see someone answer a question like this with a decent, useful answer.

  • nickandbro an hour ago

    Wow Google is becoming the new pre Llama 4 Meta when it comes to releasing open weights models.

    • embedding-shape an hour ago

      I dunno, feels a bit unfair to companies that actually do FOSS releases (Gemma 4 being released under Apache 2.0 license) to compare them to a company that never done any FOSS releases, and mostly done proprietary "available to download" releases.

      • seba_dos1 an hour ago

        Note that a binary released under Apache 2.0 license does not yet make it FOSS.

        • embedding-shape an hour ago

          Agreed, miles ahead though from "proprietary" which is what Meta been using for most model releases.

          Ideally companies would share the fucking datasets and training code already, but no, no one wants to talk about the source of those or even share the ones they have as then who knows what comes out of Pandora's box...

    • brianwawok an hour ago

      Every other Google model I have tried felt very weak compared to qwen models. I dont have a ton of use case for multimodal though, so its very possible this is a fantastic multimodal model.

      • wongarsu 28 minutes ago

        Gemma 4 27b and 32b feel pretty capable for text and visionn. Comparable with qwen, maybe a bit better on tool calling heavy tasks

        I am not overly impressed with the smaller gemma models. And gemma 3 was a bit of a mixed bag, great at some things, bad at most others

    • redman25 an hour ago

      IDK this model release is a bit disappointing considering the community has been chomping at the bit for the 124ba4b model. There was some leaked info about it but people suspect it was not released because it was too close to gemini flash in performance.

  • BiraIgnacio 30 minutes ago

    using an embedder instead of a decoder is quite clever. Not sure who came up with that first but it's a cool idea.

  • digdugdirk 36 minutes ago

    I do enjoy the immediate out of touch signaling with the "runs on your 16gb vram laptop" line. Because everyone has a laptop with 16gb vram, or can just pop out and buy a new one, right?

    • claysmithr 2 minutes ago

      I have 24 gb unified memory so it’s a good model for me

    • vehemenz 17 minutes ago

      This comment has me a bit confused.

      Consumers were complaining about the standard 8GB with the early 2020 refresh of MacBook Pros, many OSes ago. Sure, it might be workable for many tasks (as evidenced by the recent sales of the MacBook Neo), but users with a mere 8GB shouldn't have expectations of LLM performance. Even 16GB feels like a stretch.

      • NekkoDroid 7 minutes ago

        I think you are mixing up RAM and VRAM.

  • claysmithr 31 minutes ago

    I don’t see the download in lm studio

    • deckar01 2 minutes ago

      It also says it is supposed to be available in their own Edge Gallery app and it’s not there (on iOS).

  • zuminator an hour ago

    How does it compare with e4b, aside from being larger?

  • jdelman an hour ago

    I can’t help but wonder if this is the basis of the model they’ve helped tune for Apple.