Anthropic Expands Cyber Verification Program with New Access Tiers
Anthropic has revamped its Cyber Verification Program, introducing three access tiers to offer tailored AI capabilities for cybersecurity professionals.
NVIDIA Enhances MoE Training for Biological Models with Efficiency Gains
NVIDIA's BioNeMo MoE recipe optimizes training for biological foundation models, achieving 2.21x throughput gains using advanced Transformer Engine techniques.
Google Unveils Gemini 3.8 Flash TTS Models for Dynamic Voice Creation
Google launches Gemini 3.8 Flash TTS and Flash-Lite TTS, offering studio-grade voice fidelity, regional accents, and scalable audio solutions for creators and enterprises.
University of Manchester Uses NVIDIA Earth-2 for UK Air Quality Predictions
Using NVIDIA Earth-2, University of Manchester researchers deliver air pollution forecasts at unprecedented speed and resolution.
Dense vs. MoE Models: Key Tradeoffs Developers Must Know
Exploring dense vs. Mixture-of-Experts (MoE) models for AI: when to prioritize simplicity or scalability, with insights on performance and memory trade-offs.
Claude Fable 5.1 Launches in Harvey with Legal Focus
Anthropic's Claude Fable 5.1 debuts on Harvey, offering advanced AI capabilities for legal teams. Key upgrades include efficiency and improved reasoning.
Grok 4.6 Leads Biosecurity Testing on LatchBio Benchmarks
Grok 4.6 excels in biosecurity monitoring and refusal tasks, outperforming peers in LatchBio's benchmarks. Key results indicate improved safeguards.
GLM-5.3 Flash Cuts Costs by 17x with Minimal Quality Drop
GLM-5.3 Flash trims costs by 17x versus GLM-5.3 while retaining 94% task coverage, making it a cost-effective choice for coding workloads.
GLM-5.3 Beats Claude Fable 5 on DeepSWE, Costs 5.4x Less
GLM-5.3 outperforms Claude Fable 5 on DeepSWE's coding benchmark for retries and cost efficiency, at just $3.99 per rollout vs. $21.63.
GLM-5.3 Tops GPT-5.6 Sol in Cost, Edges on Multi-Try DeepSWE Tasks
GLM-5.3 offers better value than GPT-5.6 Sol for DeepSWE tasks, excelling in retry scenarios and cost efficiency, per new benchmarks.