DeepSeek's new V3.2-Exp model cuts API pricing in half to less than 3 cents per 1M input tokens

MMLU-Pro holds steady at 85.0, AIME 2025 slightly improves to 89.3, while GPQA-Diamond dips from 80.7 to 79.9. Coding and agent benchmarks tell a similar story, with Codeforces ratings rising from ...
Stark Insider

From the IT Dungeon to AI Lab: Building Stark Insider’s Research Infrastructure

For nearly two decades, Stark Insider has run on a Google Cloud VM hosting an Ubuntu server. It’s been our foundation, but ...
The BMJ

Use of new CORE risk score to predict 10 year risk of liver cirrhosis in general population: population based cohort study

Objective To develop and validate a novel risk prediction model for incident major adverse liver outcomes (MALO) in a primary care setting. Design Population based cohort study. Setting Sweden, with ...
Canadian Medical Association Journal

Cardiovascular risk thresholds for intensifying primary care encounter frequency for patients with type 2 diabetes mellitus: a target trial emulation

Background: Determining optimal timing for intensifying the frequency of physician encounters for type 2 diabetes mellitus (T2DM) requires trade-offs between timely care and clinician burden. We aimed ...
WinBuzzer

Meta Releases Code World Model as A ”Neural Debugger” Which Understands Code Logic

Meta has released Code World Model (CWM), a 32-billion-parameter AI model for researchers that simulates code execution to ...