Sign in Subscribe

Back to DP News Team

DP News Team

News by DP News Team

1,313 items

AWS maps two vLLM KV-transfer paths to separate latency bottlenecks
AWS has documented two vLLM KV-cache transfer methods: push mode moves setup off the first-token path, while bi-directional transfer avoids recomputing eligible conversation history.
Oct 20, 2026
MiniMax launches M3.1 Flash at $0.15 per million input tokens
MiniMax has released M3.1 Flash through its API, M Plan and MiniMax Code, promoting faster, lower-cost agent workloads while most performance comparisons await independent verification.
Oct 20, 2026
GMO links ARTEX to infrastructure targeting Japanese companies
GMO Cybersecurity by Ierae says it confirmed ARTEX in infrastructure targeting Japanese companies and found six AI-enabled attack hosts focused on Japan.
Oct 19, 2026
Hugging Face adds native TorchNeuron path for AWS Trainium
Hugging Face says TorchNeuron's public beta lets its major model libraries run directly on AWS Trainium, replacing the Optimum Neuron abstraction for supported workflows while retaining limits on hardware coverage and dynamic shapes.
Oct 19, 2026
OpenAI says GPT-6 models improve low-resource language coding
OpenAI unveiled MultiLangBench and a COBOL migration test, reporting gains for GPT-6 models but warning that production modernization still needs expert review.
Oct 19, 2026
Together AI launches managed post-training service for open models
Together AI has opened its managed Post-Training service, supporting supervised fine-tuning and reinforcement-learning workflows on dedicated B200 resources for selected open-weight text models.
Oct 19, 2026
AWS adds structured outputs for Claude models on Bedrock
Amazon Bedrock can now constrain Anthropic Claude responses and tool inputs to defined schemas, though AWS documents exceptions and conflicting guidance for Mantle.
Oct 19, 2026
AWS simplifies SageMaker Studio onboarding for new customers
AWS says its new-customer flow supports familiar-identity signup, automatic SageMaker Studio provisioning and Free Tier credits; existing-account sign-in is unchanged.
Oct 19, 2026
US power pipeline shifts toward gas as data-center demand rises
A report from Environmental Defense Fund and Atlas Public Policy says proposed US generating capacity is shifting toward gas amid rising data-center demand and policy changes.
Oct 19, 2026
fal launches FLUX 3 Max for three video API workflows
fal has made FLUX 3 Max available across three video-generation modes, but conflicting latency figures and a changing Elo snapshot qualify its benchmark claims.
Oct 19, 2026
MBZUAI details CT-SpatialVQA; eight tested models score below 50%
CT-SpatialVQA tests whether 3D medical vision-language models can locate anatomy and findings across CT volumes. All eight systems evaluated by its authors scored below 50% under an LLM-jury method.
Oct 19, 2026
Microsoft ends Nscale Loughton data center deal after grid delay
Microsoft ended its agreement for Nscale’s Loughton AI data center after the expected grid connection slipped into the 2030s; both companies said their other contracts continue.
Oct 19, 2026