Gemma 4 31B tops business logic benchmark

Gemma 4 31B is a paradigm shift for open-weights, proving that efficiency doesn't have to sacrifice strategic depth.

• –Its 3rd place ranking while being 20x cheaper than GPT-5.2 highlights a massive leap in instruction following and tool-use optimization.
• –A 100% survival rate in a complex 30-day simulation shows superior long-term consistency compared to much larger models like Qwen 3.5 397B.
• –The "Most Wasteful" tag (5.3% food waste) indicates a distinct "growth at all costs" bias, prioritizing revenue and reputation building over lean inventory management.
• –Native "thinking mode" integration likely provides the necessary reasoning overhead to successfully navigate multi-day dependencies and capital allocation decisions.