Infinigence AI Ranks First in 88 SOL-ExecBench GPU Kernel Tasks
中国AI刷爆英伟达GPU性能榜,这家公司拿下88项第一
- Who
- Infinigence AI
- What
- explained its agent workflow for optimizing GPU kernels on SOL-ExecBench
- Quote
下面将从容器隔离执行、Skills感知与共享Memory,以及面向不同任务的优化策略这三个方面,介绍无问芯穹具体如何支持Agent在真实硬件反馈下实现持续改进。
What happened
Infinigence AI led all participating teams with first-place results in 88 of SOL-ExecBench’s 235 GPU kernel optimization tasks, according to the article. As of October 26, it ranked first overall in the L1 and Quant tracks and second in L2 and FlashInfer-Bench. The article also describes the company’s agent system for generating, testing, and optimizing kernels against real hardware feedback.
Read next
See the full storyInfinigence AI leads 88 SOL-ExecBench GPU kernel optimization tasksThe story so far
Infinigence AI placed first in 88 of SOL-ExecBench’s 235 GPU kernel optimization tasks, leading participating teams by that measure, according to Qbit; as of October 26, it ranked first overall in the L1 and Quant tracks and second in L2 and FlashInfer-Bench. The company also explained how its agent system generates, tests, and optimizes kernels using feedback from real hardware. Its workflow uses isolated container execution, access to skills and shared memory, and optimization strategies tailored to different tasks.
Keep up with AI, one email a day
Model releases, price changes and compute developments, with original sources. Free, confirmed by email, and easy to unsubscribe. Read a past issue · 2026-10-28
