DeepMind’s Aletheia Solves Research Math as OpenAI Tests New "GPT Pro" Model
Research & Frontier Models
Google DeepMind’s Aletheia Solves Six Research-Level Math Problems: Google DeepMind’s new model, Aletheia, has successfully solved six open research-level mathematical problems using a self-correcting generator-verifier loop. The achievement highlights advancements in autonomous reasoning, supported by Google's massive 1.4 GW clean energy infrastructure investment in Minnesota.
Proprietary Model Updates
OpenAI Testing New GPT Pro Model on Web: Users have reported a new "GPT Pro" model undergoing web testing, showing significant improvements in handling complex technical tasks. The model reportedly excels at managing intricate remote desktop and Linux system configurations that previously posed challenges.
Open Source & Local Models
Qwen 3.5 Model Family Shows Strong Performance Across Scales: New iterations of the Qwen 3.5 series are demonstrating high utility, with the Unsloth-optimized 35B model proving highly effective for research and technical problem-solving. Simultaneously, the compact 4B version is being praised for its surprisingly sharp vision and image recognition capabilities on mobile devices.