Suraj Patil
@psuraj28
Followers
6K
Following
7K
Media
87
Statuses
1K
crafting pixels in the forest @bfl_ml prev: Research Engineer @pika_labs, Open Source AI @huggingface. Co-created diffusers 🧨
Paris - Kolhapur
Joined January 2016
🇺🇸 San Francisco, It's time to Hack FLUX for a weekend. See you on Nov 22-23 for a 32h hackathon. 150 spots. $25K+ in prizes. Build with FLUX models. 🌲🌲
6
13
71
three years ago, DiT replaced the legacy unet with a transformer-based denoising backbone. we knew the bulky VAEs would be the next to go -- we just waited until we could do it right. today, we introduce Representation Autoencoders (RAE). >> Retire VAEs. Use RAEs. 👇(1/n)
56
334
2K
Today we are releasing FLUX.1 Krea [dev] - a new state-of-the-art open-weights FLUX model, built for photorealism. Developed in collaboration with @krea_ai, this model is focused on images with unique aesthetics. No “AI look”, no blown-out highlights, just natural detail.
80
320
2K
*Into the land of automatic differentiation* Material is out! A short PhD course for the CS PhD in @SapienzaRoma covering basic and advanced topics in autodiff w/ slides, (rough) Notion notes, and two notebooks including a PyTorch-like implementation. 😅 https://t.co/EgMWuG7LMc
9
121
749
We've had so much fun organizing a vibe coding event for kids aged 9 to 13 that we decided to show our friends how to host one too and turn it into a global event. 📆 It’s happening on October 10-11, 2025 🧑🏫 with in-person sessions hosted locally, 🌐 all connected in one giant
20
45
200
High quality image editing no longer needs closed models We release FLUX.1 Kontext [dev] - an open weights model for proprietary-level image editing performance. Runs on consumer chips. ✓ Open weights available ✓ Best in-class performance ✓ Self-serve commercial licensing
98
361
2K
Also, I did not expect to say this, but here we are: It was a pleasure to use Overleaf again. We are publishing a report on FLUX.1 Kontext and its evaluation here: https://t.co/KpKLQSjmIw
1
1
20
We're also launching the FLUX Playground - A streamlined interface for testing our most advanced FLUX models without technical integration Available at
playground.bfl.ai
Try out the best FLUX models in your browser.
11
21
152
Try it out, the best image editing model ! The team really cooked with this one ♥️
Today we're releasing FLUX.1 Kontext - a suite of generative flow matching models that allow you to generate and edit images. Unlike traditional text-to-image models, Kontext understands both text AND images as input, enabling true in-context generation and editing.
0
0
4
no one is asking for this opinion, clearly! let the builders build! even if there is no demand the knowledge that Indian researchers will gain from training these models will only accelerate things.
Definitely not all, but much of the Indian AI scene seems like more "I want to do cool AI things that cool AI people do", not "let's solve important hard problems" No one is asking for a slightly better 24B indic model. Clearly.
2
0
3
Super cool to see Gemini Diffusion! Very curious about how it can do multi-turn generation with diffusion.
We’ve developed Gemini Diffusion: our state-of-the-art text diffusion model. Instead of predicting text directly, it learns to generate outputs by refining noise, step-by-step. This helps it excel at coding and math, where it can iterate over solutions quickly. #GoogleIO
1
0
3
Such a nice logo in the bottom right🌲Insane to see the distribution that FLUX models have reached & them becoming available on Azure. Really proud of the team.
One reservation, many models.☝️ Foundry Models now offers select partner models, with performance backed by the same SLA as Azure OpenAI. And soon, you can use provisioned throughput across models from OpenAI + xAI + DeepSeek + Meta + Mistral AI + Black Forest Labs.
4
4
80
Today, we’re announcing the preview release of ty, an extremely fast type checker and language server for Python, written in Rust. In early testing, it's 10x, 50x, even 100x faster than existing type checkers. (We've seen >600x speed-ups over Mypy in some real-world projects.)
125
507
5K
Flow Matching aims to learn a "flow" that transforms a simple source distribution (e.g. Gaussian) to an arbitrarily complex target distribution. This video shows the evolution of the marginal probability path as a source distribution is transformed to a target distribution.
25
244
2K
New blog post: let's talk about latents! https://t.co/Ddh7tXH642
sander.ai
Latent representations for generative models.
31
202
1K
Retweet if you want llama.cpp added here 👇
In preview, Copilot Pro and Copilot Free users can now bring your own key (BYOK) for popular providers such as Anthropic, Gemini, Ollama, and Open Router. This allows you to use new models that aren’t supported natively by Copilot the very first day that they’re released.
17
219
564