This iPad Pro + Thunderbolt R9700 is running Qwen3.8-27B at 158.5 tps.
Yesโฆ an iPad.
The trick
๐ฎ 32GB Radeon AI PRO R9700
๐ Thunderbolt eGPU enclosure
๐ฑ iPad Pro running LemonSeed Studio
๐ง Qwen3.8-27B Q4
๐ 131K context
And the speed progression is kind of ridiculous
๐ baseline โ 31.2 tps
โก MTP3 โ 108.7 tps
๐ DFlash2 โ 158.5 tps
Prefill:
๐ฅ 4K โ 1,636 tps
๐ฅ 32K โ 1,395 tps
LemonSeed actually runs the AMD GPU on iPadOS.
It embeds the upstream Linux amdgpu + amdkfd driver as an iPad PCIDriverKit extension, then runs its inference engine on the R9700.
Same engine for ๐ง Linux | ๐ macOS | ๐ฑ iPadOS
Same R9700 on macOS reaches 159.4 tps, so the iPadOS result is basically right there with the Mac.
โ ๏ธ Obviously, the iPad's own GPU isn't doing the 159 tps.
The iPad is hosting a 32GB desktop AMD GPU over Thunderbolt.
๐ Link in ALT.