Mistral Pixtral Large 2.5 reaches SOTA open multimodal reasoning on MMMU-Pro and MathVista-Hard
[AI-NEWS]
Date: 2026-02-25
Content : Pixtral Large 2.5 (open-weight) sets new open-model records on MMMU-Pro (visual reasoning) and MathVista-Hard (math + vision), surpassing previous open SOTA by 9–14 points while maintaining efficient inference on 8×A100 or consumer multi-GPU setups.
Keywords : multimodal reasoning, MMMU-Pro, MathVista, Pixtral Large 2.5, open-weight vision-language model
new
Meta Launches New Round of 10% Layoffs, Transfers 7,000 Employees to AI Projects
Elon Musk Loses Lawsuit Against OpenAI!
Google I/O 2026 Unveils Gemini 3.5 Series, Omni Video Generation and Spark Agent
Baidu Releases ERNIE 5.1: Pre-Training Cost Only 6% of Industry, No.1 in China Search
Kuaishou’s Kling AI Video Spinoff Planned at ~$20B Valuation
OpenAI Launches $4B Deployment Company & Acquires Tomoro for Enterprise AI
Moonshot AI's Kimi Secures $2B Funding, Valuation Exceeds $20B Post‑Investment
Musk Explains xAI Shutdown: Merged into SpaceX, Renamed SpaceXAI, Focused on Space Compute
OpenAI Launches GPT-5.5 Instant with 52.5% Lower Hallucination Rate
Microsoft and OpenAI Revised Cooperation, Ending Exclusive Tie