News
Alibaba Releases Qwen3.8-27B, an Open-Weight Multimodal Model
Alibaba’s Qwen team has released open-weight versions of its Qwen3.8 models under the Apache 2.0 license, according to The Decoder.…
News
Amazon Brings Agentic AI Extensions to Microsoft 365 Apps
Amazon has launched Microsoft 365 extensions for its Amazon Quick AI assistant, bringing agentic AI capabilities directly into Word, Excel,…
News
Anthropic Opens Public Beta for Self-Hosted Claude Code Environments
Anthropic announced a public beta of self-hosted environments for Claude Code, allowing organizations to run coding-agent sessions on their own…
News
US Army Deploys Salesforce AI Agents for 9.2 Million Soldiers and Veterans
The Pentagon and Army Human Resources Command have approved autonomous AI agents built on Salesforce’s Missionforce National Security platform to…
Engineering
Adaptive Verification in vLLM: DSpark confidence-scheduled verification
vLLM implements adaptive verification in speculative decoding using DSpark’s confidence-scored drafting mechanism, which dynamically adjusts verification length per step instead…
Opinions
How Much Longer Can China Afford Cheap AI?
Bloomberg Opinion columnist Catherine Thorbecke argues that ultra-cheap Chinese artificial intelligence, while squeezing the economics of Silicon Valley rivals, is…
Opinions
London’s DeepMind Loss Leaves a Very Bitter Taste
Bloomberg Opinion columnist Chris Hughes argues that the shift of Google DeepMind’s power base toward California underscores a broader pattern…
Engineering
Day 0 Support for Qwen3.8-2.4T-A95B on vLLM
vLLM announced Day 0 support for Qwen3.8-2.4T-A95B, a 2.4-trillion-parameter sparse mixture-of-experts model with 512 experts, running without architecture modifications. The…
Engineering
vLLM Reaches 25K Total TPS/GPU on Qwen3.5
vLLM reached 25,000 tokens per second per GPU serving Qwen3.5 on Blackwell GPUs via a disaggregated prefill-decode architecture. Three optimizations…
Engineering
LEMUR and Mean Centering for Late-Interaction Retrieval in txtai
txtai adds LEMUR and mean-centering techniques for late-interaction retrieval. LEMUR learns a fixed-dimensional encoding of multi-vector representations so late-interaction models…
Engineering
AX-Ray: Finding Causal-Leakage Defects in Two General-Purpose Public Models
VIDRAFT’s AX-Ray diagnostic framework evaluates AI models for deployment safety beyond standard capability benchmarks, using 117 structured diagnostic items across…
Engineering
Stealing Reasoning Traces from Proprietary LLM APIs
Researchers discovered a security vulnerability in proprietary LLM APIs that allowed encrypted chain-of-thought reasoning traces to be extracted and read.…