Robin Rombach

🔥 I am so damn excited to announce the launch of Black Forest Labs. We set ourselves on a mission to advance state-of-the-art, high-quality generative deep learning models for images and video, and make them available to the broadest audience possible. Today, we release FLUX.1

There is nothing more powerful than well-informed optimism. It has to be well-informed though. The "everything will be fine" type of optimism may also be somewhat useful, but it's not as useful as the "Hmm, what if we tried x?" kind.

The next generation of models won't just generate images - they'll understand worlds, motion, interaction, and action. We've been building toward this for a while. Visual intelligence is becoming real-time. @stephenbtl spoke about where we're headed at @aiDotEngineer:

@FEijkelboom @osclsd @daan_roos_ @wellingmax @mmbronstein @ismaililkanc @LucaAmb @jwvdm This is such a great blog post, congrats and thanks 🥹

Special thanks to @osclsd and @daan_roos_ for leading this work with me, and @wellingmax @mmbronstein @ismaililkanc @LucaAmb and @jwvdm for their guidance ❤️ I aim to keep this blog up to date - any feedback is welcome!

Flow-LLM Blogpost :D flow-based-llms.github.io In the last few weeks, a bunch of work on flows for language came out 🌊 That is exciting, because it makes truly parallel text generation feel real: generation where models can keep refining the whole response during inference, instead of committing token by token. I wrote an intuitive and animated introduction to the area — why autoregression has a structural ceiling, why discrete diffusion only partly escapes it, and why flows may be the first genuinely parallel alternative. Here's an overview of the key parts of the blog - and let's chat at #ICLR2026 :)

@giffmana @xeophon well... "Cohere's shareholders are set to ⁠receive around 90% of the shares in the ​combined company, whilst Aleph Alpha's shareholders will receive around ​10%, according to the report."

@xeophon Wow, this is pretty bearish on Cohere tbh

1) what

🧵 (6/N) We are witnessing a paradigm shift: Generative vision pretraining plays a foundational role similar to LLM pretraining. Image generation can serve as a universal interface for vision tasks, mirroring the role of text generation in language understanding and reasoning.

@oliver_wang2 👀

Will we ever fully replace pretrained vision encoders with generative models? Well maybe not quite yet, but it turns out they work pretty well for lots of things!

It's out, and it comes with a demo🔥 Introducing Flux.2 [klein] Move ✨ > specialized move & resize LoRA for Flux.2 [klein] 9B > draw boxes manually or pick from auto-detected objects > works with fast 4-8 step inference 💨

@fchollet @ylecun @julien_c 🌶️

@ylecun @julien_c You seem to spend a lot of time trolling on Twitter for someone who famously left Twitter

“gpt2-large is too powerful to be publicly released” vibes

@Mascobot 🫶 all the best, Marco

Life update: I've decided to leave a16z. It's been an incredible run. I joined as a Partner about 3 years ago, a very different time in AI timelines. AI was in an early takeoff stage, code hadn't reached its full potential yet, and the RL, robotics, and AI infra landscape was still taking shape. I got to work alongside some of the sharpest founders and AI researchers in tech, backing people working on the most interesting problems. It wasn't an easy decision. I've loved working with every team I've backed and worked with during this time (and I'll keep working and helping them in any way I can). Cursor had just launched in early 2023 when I met Aman co-hosting a few events at academic conferences back when AI coding was still in its early days, and it was clear since then that they were something special. Every company and team has special, from @cursor_ai, @thinkymachines, @bfl_ml, @WaveFormsAI, @deeptuneai, @ShizukuAILabs (and other unannounced ones). I'm incredibly grateful they let me be a tiny part of their journey. I'm also incredibly grateful to @bhorowitz, @martin_casado, @pmarca, @JenniferHli, and @sarahdingwang for trusting me very early on and to the entire AI infra team. This team is genuinely special and the best team founders can work with. What's next? More to come soon on this! This is something I've been thinking about for some time. The progress in AI models, RL, computer-use, robotics, has been staggering, and today we have incredibly powerful tools to work on problems that aren't yet solved and have significant potential at the current pace of progress. To the entire a16z team, thank you!

so much fun sharing the stage with @andi_blatt - founder of @bfl_ml, the company behind the flux models, and author of latent and stable diffusion - at humanx in sf to talk about visual intelligence

Releasing FLUX.2 Small Decoder: a faster, drop-in replacement for our standard decoder. → ~1.4x faster → Lower peak VRAM - decode larger images without running out of memory → Minimal quality loss → Works with FLUX.2 out of the box Especially impactful for real-time and larger resolutions pipelines.

@pallipau Missed chance for MOPS

Brainstorming how to call people like me at AI labs (i.e. generalists doing anything around growth, operations, talent, marketing, brand, comms) - Human API - MOS (member of ops staff) - MNTS (member of non technical staff) - UX to outside world What else?

@c_valenzuelab 🌶️

Hope I can still come to the show

See you there, I'll talk about @bfl_ml, what we are working on and where we are going as a research lab!

We have now updated the @CS153Systems site with our second batch of speakers including @LisaSu @satyanadella @garrytan @sdianahu @matiii and @andi_blatt Next week we'll unveil some surprises that are big wins for students and the broader ecosystem

@boyuan_chen @letz_ai @PhotaLabs Yep

Model laundering is going to be the next big integrity problem in AI. Fine-tune someone else's weights, rebrand it as proprietary, profit from the confusion. Appreciate the public call-out here. Most platforms would quietly pull it and pretend nothing happened. Naming the problem sets a standard.

After further testing of @PhotaLabs' Image "Model", we have concluded that Phota is merely a Nano Banana Wrapper that is being sold as a 'new foundation model'. This is deceptive and doesn't align with our values, nor our value proposition. We have therefore decided to withdraw this model again, after having added it earlier today. We apologize for any confusion caused, and ask PhotaLabs to adjust their communication.

@taiyasaki @sherwinbahmani Big congrats Dr @sherwinbahmani!

Congratulations Dr. @sherwinbahmani

🔥🔥🔥

@nathanbenaich @airstreet 🔥nice!

News! @airstreet has raised $232,323,232 for Fund III to back AI-first companies from the earliest stages in the US and Europe. Now the largest solo GP venture firm in Europe. Our third epoch begins today. Join us!