12 results for “distillation”
A joint evaluation by the UK Artificial Intelligence Safety Institute (UK AISI) and the US Center for AI Standards and Innovation (CAISI) has revealed. The evaluation underscores a growing concern among safety institutions: the steady.
September watch as of 9 Sep: Daybreak keys and early drops still stand; mid-month added Muse consumer reporting, OpenAI Images/Opal/research, and cyber sandbox escapes. Gemini 4 and Apple China launch timing remain open.
Two dozen companies and organizations have signed an open letter urging United States policymakers to protect open-weight artificial intelligence models. The letter carves out space for one technique that has become contentious in AI.
Runway announced Solaris, an Interface World Model that generates interactive UIs frame by frame at 720p with 37 ms latency, removing the code layer between design and running app. The shift challenges coding assistants and design tools by making the model itself responsible for every pixel and interaction.
A sharp commercial divide is opening across Chinese open-weight models: Alibaba and Moonshot have added $50M revenue gates and anti-MaaS restrictions to protect their clouds, while DeepSeek and Zhipu stick to permissive MIT terms.
Anthropic released Claude Fable 5.1 for general availability and Mythos 5.1 through trusted access. Benchmarks jump on Terminal-Bench-Science and agentic coding; cache-read pricing drops 75% for roughly 25–45% lower total cost at unchanged $10/$50 input/output rates.
Tencent open-sourced Hy-MT1.5-1.8B packs that shrink a 3.3 GB translation model to as little as about 440 MB for offline use on ordinary phones. The 1.25-bit and 2-bit releases keep 33-language coverage while targeting on-device privacy and real-time translation without a network.
A developer known as slvDev has demonstrated a 28.9-million-parameter language model generating text at 9.88 tokens per second on an ESP32-S3. Running a language model on a sub-$10 microcontroller changes the cost floor for local AI.
LTX, the open world model company spun out of Lightricks, has released LTX-2.5 — a rebuilt video and world model that generates a 10-second, 720p clip. The release reinforces a business model bet that open weights — not closed APIs — will.
Z.ai released GLM-5.2, a 753B open-weight model with MIT licensing that matches Anthropic's Opus 4.8 on agentic benchmarks at roughly one-fifth the token cost. The 1M-context model is already running in Cursor and Claude Code, shifting enterprise leverage toward self-hosted inference.
Liquid AI has released the LFM2.5 model family, a comprehensive suite of open-weight models designed to run entirely on edge devices without cloud. The LFM2.5 releases address three critical constraints that have limited on-device AI.
Koray Kavukcuoglu founded DeepMind’s deep-learning team and helped pioneer DQN and WaveNet. This career profile covers his ladder to SVP and Chief AI Architect — separate from the August 2026 org news.