Which GPUs does FlashAttention-3 support?

FlashAttention-3 main (2026-09-27): 2 of the 65 GPUs tracked here are supported, from the library's own requirements. Checked 2026-09-27.

The rule

The README's stated requirement is "H100 / H800 GPU, CUDA >= 12.3". FlashAttention-3 is built from the repo's hopper/ directory and is Hopper-only. FlashAttention-4 (the flash-attn-4 package, 4.0.0b32 on 2026-09-27) is a beta and is not tracked here yet.

Supported (2)

GPUVendorTargetDetails
H200 SXM 141GBNVIDIA9.0
H100 SXM5 80GBNVIDIA9.0

Not supported (62)

GPUVendorTargetDetails
B300 (Blackwell Ultra)NVIDIA10.3no build for compute capability 10.3
DGX SparkNVIDIA12.1no build for compute capability 12.1
RTX 5060 Ti 16GBNVIDIA12.0no build for compute capability 12.0
RTX 5090NVIDIA12.0no build for compute capability 12.0
RTX PRO 6000 BlackwellNVIDIA12.0no build for compute capability 12.0
B200NVIDIA10.0no build for compute capability 10.0
L40SNVIDIA8.9no build for compute capability 8.9
RTX 4060 Ti 16GBNVIDIA8.9no build for compute capability 8.9
RTX 4090NVIDIA8.9no build for compute capability 8.9
RTX 6000 Ada GenerationNVIDIA8.9no build for compute capability 8.9
RTX 3060 12GBNVIDIA8.6no build for compute capability 8.6
A100 SXM4 80GBNVIDIA8.0no build for compute capability 8.0
RTX 3090NVIDIA8.6no build for compute capability 8.6
T4NVIDIA7.5no build for compute capability 7.5
TITAN RTXNVIDIA7.5no build for compute capability 7.5
GTX 1080 TiNVIDIA6.1no build for compute capability 6.1
TITAN VNVIDIA7.0no build for compute capability 7.0
V100 SXM2 32GBNVIDIA7.0no build for compute capability 7.0
P100 SXM2 16GBNVIDIA6.0no build for compute capability 6.0
Tesla P40NVIDIA6.1no build for compute capability 6.1
TITAN X (Pascal)NVIDIA6.1no build for compute capability 6.1
M40 24GBNVIDIA5.2no build for compute capability 5.2
K80NVIDIA3.7no build for compute capability 3.7
Instinct MI350XAMDgfx950
Instinct MI355XAMDgfx950
Radeon AI PRO R9700AMDgfx1201
Radeon RX 9060 XT (16GB)AMDgfx1200
Radeon RX 9060 XT (8GB)AMDgfx1200
Radeon RX 9070AMDgfx1201
Radeon RX 9070 XTAMDgfx1201
Ryzen AI Max+ 395AMDgfx1151
Instinct MI325XAMDgfx942
Radeon RX 7600 XTAMDgfx1102
Instinct MI300AAMDgfx942
Instinct MI300XAMDgfx942
Radeon RX 7700 XTAMDgfx1101
Radeon RX 7800 XTAMDgfx1101
Radeon RX 7900 GREAMDgfx1100
Instinct MI210AMDgfx90a
Radeon RX 6650 XTAMDgfx1032
Radeon RX 6750 XTAMDgfx1031
Radeon RX 6950 XTAMDgfx1030
Radeon RX 7900 XTAMDgfx1100
Radeon RX 7900 XTXAMDgfx1100
Instinct MI250XAMDgfx90a
Radeon RX 6600AMDgfx1032
Radeon RX 6700 XTAMDgfx1031
Instinct MI100AMDgfx908
Radeon RX 6800AMDgfx1030
Radeon RX 6800 XTAMDgfx1030
Radeon RX 6900 XTAMDgfx1030
Radeon VIIAMDgfx906
Radeon Instinct MI50AMDgfx906
Radeon Instinct MI60AMDgfx906
Radeon Instinct MI25AMDgfx900
Radeon Instinct MI6AMDgfx803
Radeon Instinct MI8AMDgfx803
Arc B580Intel—
Data Center GPU Max 1550Intel—
Arc A380Intel—
Arc A750Intel—
Arc A770 (16GB)Intel—

Not enough data (1)

GPUVendorTargetDetails
Instinct MI455XAMD—gfx target not recorded

Sources: github.com/Dao-AILab/flash-attention/blob/main/README.md

← All libraries, every GPU · Install checker