Added lots of improvements to the LLM Architecture Gallery in the last 2 weeks. Imho the coolest one yet: A diff tool many of you were asking for! sebastianraschka.com/llm-arc…
GENERATIVE AI
-
Mistral Launches Voxtral TTS Model Outperforming ElevenLabs
By
–
Mistral just dropped Voxtral TTS an open-weight text-to-speech model with ultra-low latency, emotional expressiveness, and support for 9 languages. It outperformed ElevenLabs v2.5 Flash in zero-shot voice cloning tests judged by native speakers. pic.twitter.com/GaNGRC4O15
— Chubby♨️ (@kimmonismus) 26 mars 2026Mistral just dropped Voxtral TTS an open-weight text-to-speech model with ultra-low latency, emotional expressiveness, and support for 9 languages. It outperformed ElevenLabs v2.5 Flash in zero-shot voice cloning tests judged by native speakers.
-
Pika AI Selves Public Beta Launch: Personal AI Agents Available
By
–
WE INTERRUPT YOUR REGULAR PROGRAMMING to say that Pika AI Selves are officially in public beta!
— Pika (@pika_labs) 26 mars 2026
Everyone can now give birth to an agentic extension of themselves at Pika dot me, or on our new iOS app. pic.twitter.com/Qce0CuyZ0mWE INTERRUPT YOUR REGULAR PROGRAMMING to say that Pika AI Selves are officially in public beta! Everyone can now give birth to an agentic extension of themselves at Pika dot me, or on our new iOS app.
-
Single Creator Builds Entire Cyberpunk World with AI Gaussian Splats
By
–
AI can generate 100 million splats, but it’s one creator’s imagination that made this uniquely beautiful world possible!🤩 https://t.co/3yIemVLLPH
— Fei-Fei Li (@drfeifei) 26 mars 2026AI can generate 100 million splats, but it’s one creator’s imagination that made this uniquely beautiful world possible!🤩 World Labs (@theworldlabs) This entire cyberpunk world was built by a single creator. 100 million Gaussian splats, with nearly every surface and structure generated in Marble. We’re entering an era where individuals can build entire worlds. — https://nitter.net/theworldlabs/status/2037200344267227301#m
-
AI Replication Kills Business Moats Except Taste
By
–
>This guys build a possible $1M business
— Linus ✦ Ekenstam (@LinusEkenstam) 26 mars 2026
>Another dude in his comments runs 3 prompts with AI
>Makes a replica in a few minutes (90% there)
>OpenSource it in Github
>There is ZERO moat left
>Except Taste 👅 https://t.co/HhVoy6hU4H>This guys build a possible $1M business >Another dude in his comments runs 3 prompts with AI >Makes a replica in a few minutes (90% there) >OpenSource it in Github >There is ZERO moat left >Except Taste
-
Low Latency Responses Create More Human-Like Conversations
By
–
The faster response is big, feels more human. The lower the latency, the more natural the conversation. pic.twitter.com/nGEiXFa5Ht
— Chubby♨️ (@kimmonismus) 26 mars 2026The faster response is big, feels more human. The lower the latency, the more natural the conversation.
-
Composer 2 Real-Time RL Enables Model Updates Every Five Hours
By
–
Earlier this week, we published our technical report on Composer 2. We're sharing additional research on how we train new checkpoints. With real-time RL, we can ship improved versions of the model every five hours.
-
AI Service Pricing Strategy During Waitlist Phase
By
–
we are figuring it out! cost to run it is high, but we want to make it as accessible as we can. will include it in the base @every subscription for now while we get people off the waitlist and follow up quickly with permanent pricing!
-
Voxtral TTS: Open-source, fast, 9 languages
By
–
Voxtral TTS by @MistralAI is officially out! Open-source and fast TTS model with a support of 9 languages! https://t.co/EDdhpR8J6g pic.twitter.com/5o5p5aEYRk
— 🚨 AI News | TestingCatalog (@testingcatalog) 26 mars 2026Voxtral TTS by @MistralAI is officially out! Open-source and fast TTS model with a support of 9 languages!
-
Gemini 3.1 Flash Live Launches with Native Audio Understanding
By
–
📢 Another exciting step forward today with the launch of Gemini 3.1 Flash Live. It natively understands audio, making it much more capable of handling complex instructions. It leads on ComplexFuncBench, and on Scale AI’s AudioMultiChallenge, demonstrating skill in complex instruction following and long-horizon reasoning amidst the interruptions of real-world audio. That means the model can pick up on nuances like pitch and pace, leading to much more fluid, high-fidelity voice interactions. It’s now powering Gemini Live and Search Live globally. Huge congrats to the teams that made this possible. ✨ blog.google/innovation-and-a…