MiniMax, a Chinese AI Company, Unveils ‘M2.5’: A Cost-Effective Alternative to GPT, Gemini, and Claude with Comparable Performance

On February 13, 2026, MiniMax, a Shanghai-based AI firm, launched the MiniMax M2.5, a groundbreaking model designed to enhance productivity dramatically in real-world applications. Utilizing advanced reinforcement learning across hundreds of thousands of machines, the M2.5 showcases exceptional processing speed and cost-efficiency. Notable improvements over its predecessor, the M2.1, include achieving 80.2% in the SWE-Bench Verified benchmark and supporting over 10 programming languages, such as Python and Java. The M2.5 is engineered for comprehensive development lifecycles, providing capabilities in full-stack projects for platforms like web, Android, and iOS. With a self-developed reinforcement learning framework named ‘Forge’, it significantly accelerates learning, yielding a 40x increase in speed. Available in two models—M2.5 and M2.5-Lightning—this AI tool is 10 to 20 times more cost-effective than other leading models, allowing businesses to leverage advanced AI services without financial strain. For further information, visit MiniMax’s official site.

Source link