Meet Fibo 2.0 from @bria_ai_ 🔥 The #1# *open weights* model in image editing is supported in Diffusers (day-0, of course)! 1. JSON-native control 2. Pretty good text rendering 3. Comes in two variants: Turbo and Dev Check out the model here:

Was meant to be a Valentine's gift, but I'm releasing it early ❤️ Wrote a long piece on KV caching in flow models for image generation. Includes intuition, commentary, implementation details, benchmarks, & some quirky experiments.

Tensor parallel loading in 🤗 Diffusers got a massive upgrade 🔥 On Flux.2-Dev DiT, with a TP degree of 4 (A10G): • 30.4s → 12.5s loading • 64.1 → 6.8 GB peak CPU memory/rank ~2.4× faster loading. ~89% less CPU memory. Check it out:

Here's a thread on how KV Caching is incorporated in QwenImage 2.1 👇 It separates the fixed context from image positions that change at every denoising step. Cache the fixed part once, and later steps get much cheaper. We measured a 2.55x speedup 💥 1/8

Hot off the press on a Sunday! Available Day-0 in Diffusers 🤗 The model is really cool and has an interesting way of using KV-caching and a disentangled prefill + decode. Will do a separate thread on that! Check out the model here:

Found a faster kernel? You shouldn’t need to rewrite your model to use it. With 🤗 Kernels, you can choose which kernel runs a supported layer and replace its forward() with an optimized implementation. Here’s how 🧵 1/5

We presented our work Flash-BoN on the 11th Sept at #ECCV26#. It was a fulfilling experience, to say the least! @RawalRuchit told me about the idea in Hawai'i during ICCV'25, and I was immediately like, let's go! The origins of the work started with a curiosity: Change the metric for inference-time scaling algos in diffusion from number of function evals (NFE) to something more bounded, like wall-clock time. We decided to spend that (inference-time) precious compute exploring more candidates WITHOUT busting the tanks. It's so cool that Flash-BoN works across the board: T2I, T2V; different model scales; complements other techniques like BFS, ReflectionFlow; and even improves the convergence of Flow-GRPO. If you haven't checked it out yet, here's the link:

The only tutorial you should attend at #ECCV26# (kidding 😂). But I am incredibly psyched to be doing this with the best bunch out there. I strongly feel this is a timely tutorial! We will present general approaches alongside our learnings from (post)-training impactful models like Flux2/3. Tutorial website: Save your calendars!

One plausible future is not enough. A video model may generate a convincing die roll, yet produce the same few outcomes again and again. Our new paper asks a simple question: Does the model capture not only what can happen, but how often? Introducing PAWBench 🧵

We released Diffusers 0.40.0 🔥 Main highlights for me: > Graduating Modular Diffusers out of experimental. > Bunch of new really cool models as always: LTX2.5, MiniMax H3, Music 3, Wan Animate 2, etc. > Tensor-parallel support for a select few models. > New quantization backends: SDNQ, Nunchaku-Lite. > MASSIVE redesign of our tests. It should now be leaner and better. Sometimes faster, too. Check out the release notes here:

What a week for the open-weights community! The Diffusers team, hence, has been extremely busy at work 👨‍💻 > MiniMax H3: (video + audio) > LTX2.5: (video + audio) > Wan Animate 2: > MiniMax Music 3: Some of the most intellectually satisfying stuff, I must say! If you haven't got the chance to play with them yet, the coming weekend is a good time!

We shipped a new repo type called "kernel" on the Hub. We want to democratize the whole ping-pong around packaging, distributing, and using custom kernels. This repo type is only available to a few community partners, @sgl_project being the first! Hop in 🧵for more details.