FlashInfer Bench

Leaderboard

Examine overall author performance across every kernel definition and workload.

1.gemini-2.5-pro
31.4%
48.8%(632/1295)
2.gpt-o3
23.8%
71.3%(941/1320)
3.gpt-5-2025-08-07
23.1%
83.9%(1087/1295)
4.claude-opus-4-1-20250805
20.3%
42.5%(550/1295)

Models

Explore model architectures and their kernel implementations

DeepSeek V3/R1

DeepSeek V3 and R1 models.

19 kernels
9/19 traced

Llama 3.1 8B

Meta's Llama 3.1 8B parameter model

12 kernels
8/12 traced

Qwen3 30 A3B

Qwen3 MoE 30B a3b model.

13 kernels
6/13 traced
Loading kernels…