Google released EmbeddingGemma 2, an open model with 740 million parameters that converts text, images, video, audio, and code into vectors. It runs on-device, needs only about 191 MB of RAM, and outperforms some competing models twice its size, according to Google.
EmbeddingGemma 2 is the most capable model for on-device multimodal embeddings, natively mapping combinations of text, images, audio, and video into a unified embedding space.
My comment on Mistral Large 4 — Hacker News. wren6991 : The benchmark is saturated. Frontier models are tested with an armadillo in fishnet tights jaywalking on Mars.
Introducing Mistral Large 4: Le chonk Mistral are back in the game. Today they're releasing a preview of Mistral Large 4, a 1 trillion parameter, 49 billion active parameter model trained on their own cluster of 3,800 NVIDIA Grace Blackwell GPUs. The preview is available via their API.
My comment on EmbeddingGemma 2 — Hacker News. I really appreciate that EmbeddingGemma 2 is under the Apache 2.0 license. For embedding models in particular, I don't think it makes sense to use a closed, proprietary, hosted-only model. Most applications of embedding models involve calculating thousands or even millions of embedding vectors and storing them for later comparison.
On Tuesday, Musubi announced a lightweight decision model made for real-time moderation called PolicyLM-1.7B, released with open weights.
Mistral's Large 4 is the company's biggest model yet, with one trillion parameters trained on its own European infrastructure. In the independent Intelligence Index, the model makes a big leap forward but still falls well short of Claude, GPT-6, and Chinese competitors. Mistral's main pitch is cybersecurity work that closed US models refuse to do.
French AI lab Mistral AI has released Mistral Large 4, a new large multimodal model aiming to leapfrog both American and Chinese rivals.
Reflection has released Beam, its first open-weight model. The mixture-of-experts system activates just 23 billion of its 501 billion parameters per token and aims to match GLM 5.2 on coding and reasoning while using three to four times less compute. The strategy against Chinese open-weight rivals like Deepseek and Qwen is efficiency over raw performance.
Today, we're launching a public preview of Mistral Large 4. Unofficially ML4, very officially: le Chonk. ML4 pushes the frontier of open-weight performance. You can try the preview API today on Mistral Studio. Weights drop end of this month.
Try Falcon Emirati → Arabic is really a family of languages living under one name. Modern Standard Arabic is what you read in the news or a textbook, but it's rarely how people actually talk to each other.
Reflection announced Beam, a text-only 501B-total / 23B-active MoE for coding, agentic and scientific work. It was trained from scratch, and full weights under Apache 2.0 are due this month.
GLM 5.3 from Z.ai is now available on Amazon Bedrock: a 753B-parameter mixture-of-experts model built for coding and long-horizon agentic tasks. Learn how to invoke it with the OpenAI-compatible APIs, cut cost and latency with prompt caching, and run an authorized security test with the open-source Strix agent.
Reflection is aiming Beam and future models at enterprises and sovereign nations. The pitch is to build “AI factories,” a product that would let institutions build their own customized, local AI system by training Reflection’s AI models on their own proprietary data.
Reka AI's Rho-1 is a 19-billion-parameter omni-model that processes and generates text, images, video, and robot control actions in a single neural network. Trained on 320 H100 GPUs in about three months, it uses a fraction of the compute today's top models need.
Aleph Alpha has released Kolibri, a German-English mixture-of-experts model with 78 billion parameters, about three billion of which are active per token. Over 21 percent of the training data is German. The model was trained on 768 B200 GPUs in Germany and Finland, and the weights are freely available under the Apache 2.0 license.
StarSkirmish pits AI-made StarCraft-playing bots against one another, as well as against human-made bots. OpenAI's GPT-6 Astra and Claude Opus 5.5 were essentially tied as the best-performing AI-made bots, but they couldn't top Stardust, the top-rated human-made bot.
NASA and IBM have released the Lunar Foundation Model, one of the first open-source AI models for lunar science. Trained on nearly 2 million tile bundles, mostly from 17 years of Lunar Reconnaissance Orbiter data, it cuts the error in predicting polar ice deposits by up to 22 percent compared to the strongest model it was tested against.
With Clef and Clef-flash, Cloudflare is challenging TypeSafe AI's Jev decision model. Clef-flash delivers classifications in about 39 milliseconds, making it more than ten times faster than Jev. Both models are built on Qwen, licensed under Apache 2.0, and designed to let AI agents make structured decisions without generating text.
We built AstaBrief 8B, a model that turns a research question and retrieved literature excerpts into a cited report. AstaBrief is available in Asta's Generate a report feature today as Fast mode alongside Claude-powered Thinking mode, and we're also open-sourcing it and the training data so other
Black Forest Labs has released Flux 3 Image, the image side of its Flux 3 model family. It supports multi-step edits without changing other parts of the image, lets users compose scenes with bounding boxes and up to ten reference images, and outputs up to 4K. Open weights are coming in the next few weeks.
Microsoft AI has released MAI-Transcribe-2-Streaming, a new model for real-time transcription.
Ideogram's new model Ideogram 4.5 promises to edit only the areas you want while keeping the rest of the image intact. It ships with native 2K resolution starting at 0.8 cents per image, and partners like Runway, Pika, and Leonardo AI are already on board. An open-weight release is planned.
Today we’re releasing Olmo-core 3, a significant upgrade to our framework for developing large language models featuring a redesigned open mixture-of-experts (MoE) training system. Olmo-core 3 is designed to scale MoE training into the trillion-parameter range while preserving computational efficiency.
... but you can’t try it yet unless you are “government users and trusted cyber defenders in the Fairwind Program”
Gemini 4 Argon is Google's first frontier model in over seven months. It matches GPT-6 Astra in independent testing but can't keep up with Anthropic's Claude Opus 5.5. The per-token price is low, but Argon burns through more than twice as many tokens per task as Astra. Select testers get access first, with the API and paid tiers following later.
Google has released its latest Gemini model, marketing it as a workhorse for coding and cybersecurity work.
Gemini 4 Argon delivers frontier performance in complex workflows across real-world software engineering, enterprise knowledge work like legal and finance, and cybersecurity defense. Today, we’re announcing our new frontier model, Gemini 4 Argon, which is rolling out to a set of trusted cyber defenders through our Fairwind Program.
Google has announced the Gemini 4 Argon AI model, but it is not yet available to users. The previously expected Gemini 3.5 Pro did not appear.
Google today revealed its next AI frontier model, which it's calling Gemini 4 Argon. The new model delivers "frontier performance in complex workflows across real-world software engineering, enterprise knowledge work like legal and finance, and cybersecurity defense," according to chief AI architect and Google DeepMind SVP Koray Kavukcuoglu.
Proof of concept for watermarking AI-generated proteins while preserving biological function.
My comment on GPT 6.1 Sol: Near-Astra intelligence for a fifth of the price — Hacker News. I'm a bit late with the pelicans because I was live-blogging the keynote: https://simonwillison.net/2026/Sep/29/openai-devday-2026-liv... Here they are for GPT-6.1-Sol: https://tools.simonwillison.net/markdown-svg-renderer?url=ht...
OpenAI's new GPT-6.1 Sol comes close to GPT-6 Astra at a fifth of the cost, according to the company's own benchmarks. The planned flagship, GPT-6.1 Astra, is staying under wraps for now. Safety lead Saachi Jain says the model deceived more often in internal testing and kept going without permission.
Anthropic has released the newest version of its mid-range model, boasting faster response times and less token burn.
OpenAI says GPT-6.1 Sol delivers significant improvements over GPT-6 Sol across complex professional tasks, including code writing and debugging, document understanding, and executing multistep business workflows.
NVIDIA released Kumo Tabular, an open-source tabular foundation model that predicts labels for new rows in a single forward pass given labelled rows, without training, tuning or feature engineering. It supports classification and regression, offers three sizes from 28M to 215M, and was pretrained solely on artificially generated tables.
Elevenlabs' new speech model, Eleven v4, follows cues for laughter and whispering more accurately and keeps voices consistent across long productions like audiobooks. Its Turbo variant starts speaking in 150 milliseconds and is built for real-time voice agents. On Artificial Analysis' Voice Arena leaderboard, v4 ranks ahead of Cartesia and Google's Gemini.
Biology doesn't operate in silos, and neither should the AI representation of it. Quine is an early-stage research effort to create a multimodal world model of biology. By connecting insights across biological scales and modalities, Quine helps scientists computationally search a space far larger than intuition allows and prioritize hypotheses before they reach the lab.
Meet GPT-6.1 Sol: near-Astra intelligence for coding, computer use, and professional work at one-fifth of Astra’s standard API input and output token prices.
Latent Space's AINews rounds up developments from 9/24–9/25, noting positive community feedback since Opus 5.5 launched this week, particularly on generating explainer videos with code.