Ivan Lobov @ilobov

Those folks ship!

Gradium @GradiumAI

4 days ago

Today we launch stt-translate and s2s-translate: real-time speech-to-text and speech-to-speech translation. They compete with gemini-3.5-live-translate and gpt-realtime-translate on latency and quality, while allowing you to speak in any voice from our catalog or one you clone.

0 8 34 9K 19

@_inception_ai Wow, what a cheap shot, folks! In order to claim "Pareto frontier for quality, speed, and cost" in diffusion model, one would actually need to show Mercury 2 speed and cost. How large is Mercury 2? How many GPUs does it run it? Then we could properly compare.

lol if true. I mean, why not at least train on a new persona? Although I suspect this is fake.

Cooper @peakcooper

2 weeks ago

GLM 5.2 is absolutely convinced that it is actually Claude, from Anthropic. When I tell it that it's GLM 5.2, it refuses to believe me, but is willing to check the local agent config to see what model is running. The realization:

keyboard_arrow_left Previous keyboard_arrow_right Next

321 244 5K 587K 1K

GLM 5.2 is absolutely convinced that it is actually Claude, from Anthropic. When I tell it that it's GLM 5.2, it refuses to believe me, but is willing to check the local agent config to see what model is running. The realization:

keyboard_arrow_left Previous keyboard_arrow_right Next

You've been building it for several years now. You are close / at frontier and you know exactly how hard it is - from compute to infra to training recipe. And you believe that, say, if somebody would start the same journey today they would have a fighting chance? Or, say, they would have the right talent, but no 40k gb300 readily available. Would they have a fighting chance?

It is quite sad, but I do agree with this essay - anyone who does not have access to a frontier-level intelligence this year, is unlikely to catch up with it. Having worked a bit with AI hardware and (close to) frontier-level models, it is obvious to me that the amount of investment, talent and infrastructure required is just beyond what most actors on the planet can muster in a short amount of time. And truly frontier-level models + harnesses give an enormous advantage over competitors.

Andrew Curran @AndrewCurran_

2 weeks ago

x.com/i/article/2066…

260 647 4K 2.4M 5K

And this startup list only covers the ML and software side of things. If the US decides to export control hardware to Europe, Canada, Japan, etc. What are we going to do then? Let's just hope that uncle Jensen has a good enough lobby in the white house.

Nando de Freitas @NandoDF

2 weeks ago

No one should be surprised by this. The USA is doing what any self-interested nation state would do. The real question is why are Europe, Canada, Australia, Korea, Japan and UK not able to compete seriously. That is the question everyone in government needs to answer. And no,

35 42 402 95K 144

Wow. Interesting how this decision will age...

Anthropic @AnthropicAI

2 weeks ago

The US government, citing national security authorities, has issued an export control directive to suspend all access to Fable 5 and Mythos 5 by any foreign national, whether inside or outside the United States, including foreign national Anthropic employees. The net effect of

13K 26K 88K 92.1M 24K

The US government, citing national security authorities, has issued an export control directive to suspend all access to Fable 5 and Mythos 5 by any foreign national, whether inside or outside the United States, including foreign national Anthropic employees. The net effect of this order is that we must abruptly disable Fable 5 and Mythos 5 for all our customers to ensure compliance. Access to all other Claude models is not affected. We apologize for this disruption to our customers. We believe this is a misunderstanding and are working to restore access as soon as possible. Read our full statement: anthropic.com/news/fable-myt…

@volokuleshov @mariannearr Apologies for nitpicking, but we do not use encoder-decoder architecture. Our encoder is the same model as the decoder. Having said that, you are doing great research, awesome work!

DiffusionGemma is great at tweaking to iterate 🔥 fast ⚡️ watch it generate and tweak a website frontend ⤵️ this is simple but imagine the possibilities 🤯

+1, what a d*ck move, imho. Not an unbiased opinion, but makes one appreciate open-weights even more...

bayes @bayeslord

3 weeks ago

They didn’t mean pause AI research, they meant pause *your* AI research

50 343 5K 105K 167

They didn’t mean pause AI research, they meant pause *your* AI research

@Nate_Keating Are we keeping it for the summer, do you know? Still haven't gotten a chance to play with it. 😭

@MaxBrinAI I would recommend blogpost from @MaartenGr for details. He did an awesome job describing f the differences newsletter.maartengrootendorst.com/p/a-visual-gui…

Our team just launched a new SoTA open-weights text diffusion model. It's been a wild year here, folks!

Google DeepMind @GoogleDeepMind

3 weeks ago

DiffusionGemma is our new experimental open model with up to 4x faster output on dedicated GPUs. Instead of predicting word-by-word, it generates entire blocks of text simultaneously. This lets the model self-correct and format complex markdown in real time.

110 269 2K 1,000K 484

Auto regressive LLMs are officially on notice. run Gemma 4 26B diffusion gguf with llama.cpp Google just dropped DiffusionGemma-26B, and it completely flips how we generate text. instead of predicting words one by one, it generates 256 tokens in parallel using bi-directional attention. its like stable diffusion, but for language. the model starts with random text "noise" and iteratively refines and self-corrects the entire block in real-time to fix formatting and reasoning errors on the fly. since it’s a Mixture of Experts (MoE) that only activates 3.8B parameters during inference, it fits perfectly on consumer hardware. You can run the Q4_K_M quant with an 18GB VRAM budget on a single RTX 3090 or RTX 4090 with exceptional throughput. Tested on Ubuntu 22 with CUDA 13.1 using the cutting edge experimental llama.cpp branch. Here is how to compile and run it with the live terminal denoising visualizer: # 1. Clone & check out the experimental PR (#24423) - 1) git clone github.com/ggml-org/llama… && cd llama.cpp -git fetch origin 2) pull/24423/head:diffusiongemma && --git checkout diffusiongemma # 2. Build with CUDA support 1) cmake -B build -DGGML_CUDA=ON -DCMAKE_CUDA_ARCHITECTURES=native 2) cmake --build build -j $(nproc) --config Release --target llama-diffusion-cli # 3. Run with live visual denoising (llama.cpp flags) ./build/bin/llama-diffusion-cli \ -m /path/to/diffusiongemma-26B-A4B-it-Q4_K_M.gguf \ -ngl 99 -cnv -n 2048 --diffusion-visual Watch the video below to see the live --diffusion-visual canvas iteratively de noising the prompt output in real time. guide and unsloth's hugging face GGUF model links are in the comments below! Is auto regressive generation officially legacy tech? Let me know what you think.

Google Gemma @googlegemma

3 weeks ago

Meet DiffusionGemma! An experimental open model that explores a fast approach to text generation, released under an Apache 2.0 license. Moving beyond sequential, token-by-token processes to generate entire blocks of text simultaneously. Here’s what’s new with DiffusionGemma: 👇

166 809 5K 958K 2K

@vanstriendaniel @googlegemma That's really cool. 10 dns, is it with thinking enabled? I would expect it to be even fewer dns on this task, but without thinking.

Can @googlegemma DiffusionGemma help fix broken OCR? In theory, denoising tokens in parallel could work better for OCR correction since context is seen upfront? Pointed it at 19th-century newspaper OCR. It corrected better than the autoregressive baseline — at ~8x the speed.

Interesting, it seems that standard methods of making a model unsafe do not work on DiffusionGemma! Maybe an unexpected feature of the technology? huggingface.co/DuoNeural/diff…

DiffusionGemma brings high intelligence and lightning fast ⚡️ inference to local developers (>1100 tok/s on a single H100)! I'm excited to see what people will do with this model - and what improvements people can build on top (better samplers maybe??). So unbelievably proud of the hard work the team put in to get this out!🪐🪐🪐

Sundar Pichai @sundarpichai

3 weeks ago

DiffusionGemma is an open, experimental model that brings our text diffusion research to Gemma 4. It’s a racehorse 🏇achieving up to 4x faster inference by generating entire blocks of text simultaneously vs predicting token-by-token (word-by-word) output!

193 408 3K 312K 543