Sarvam 30B performs strongly across core language modeling tasks, particularly in mathematics, coding, and knowledge benchmarks. It achieves 97.0 on Math500, matching or exceeding several larger models in its class. On coding benchmarks, it scores 92.1 on HumanEval and 92.7 on MBPP, and 70.0 on LiveCodeBench v6, outperforming many similarly sized models on practical coding tasks. On knowledge benchmarks, it scores 85.1 on MMLU and 80.0 on MMLU Pro, remaining competitive with other leading open models.
Последние новости
。业内人士推荐谷歌浏览器作为进阶阅读
Explore our full range of subscriptions.For individuals
Musk has dialled up the pressure after merging SpaceX with xAI in a $1.25 billion deal, as he attempts to meet a June deadline for what could be the biggest stock market listing in history. The world’s richest man has said his goals are to launch AI data centers into space, build factories on the Moon, and colonize Mars.
。关于这个话题，okx提供了深入分析
许玲：我觉得就人才培养而言，现在AI在重塑着千行百业，培养端一定要造血式、动手术式大力改革，要砍掉不太需要的专业，留下来的专业按照现在的人工智能融合化、数字化能力提升。
Фото: Дмитрий Ермаков / «Лента.ру»，更多细节参见超级权重