Qwen3.8-27B: 159 tok/s on R9700, 64 tok/s on Strix Halo
LemonSeed Studio, an iPad editor/IDE, demonstrates on-device inference using an AMD GPU in a Thunderbolt enclosure. It utilizes an unmodified Linux amdgpu + amdkfd driver as a PCIDriverKit extension to run LemonSeed Engine (LSE) on the GPU. Performance tests with Qwen3.8-27B Q4 show impressive speeds: 159.4 tok/s on an R9700 with macOS, 158.5 tok/s on iPadOS, and 64.4 tok/s on a Strix Halo (Radeon 8060S) with Linux, highlighting efficient GPU utilization for AI tasks.
Time & source
Times shown in UTC
Display time zone: UTC
Local time zone unavailable; showing UTC.
PublishedOffset at this time: UTC+0Oct 9, 2026, 01:20 UTC
IngestedOffset at this time: UTC+0Oct 9, 2026, 09:00 UTC
- Published
- Oct 9, 2026, 01:20
- Ingested
- Oct 9, 2026, 09:00
- Source type
- Dev community
- Tier
- Community
- Source status
- Healthy
Tier is a per-source editorial setting, not a per-item score.
LemonSeed Studio is an iPad editor/IDE with on-device inference on an AMD GPU in a Thunderbolt enclosure. It embeds the unmodified upstream Linux amdgpu + amdkfd driver (mac_linuxgpu) as a PCIDriverKit extension, and runs LemonSeed Engine (LSE) on the GPU. In the photos: iPad Pro + Sapphire Radeon AI PRO R9700, Qwen3.8-27B Q4 with a Q8 DFlash2 draft, 131k context.
LSE is the same engine on every platform: Linux, macOS and iPadOS. It records the model's forward pass as a graph, fuses ops and generates kernels for the GPU it's on. Where a kernel has several layouts, it measures each on the device and keeps the fastest. Speculative decoding (MTP and DFlash2) picks how many draft tokens to verify from measured acceptance and cost.
Decode, code prompt, Qwen3.8-27B Q4 (baseline / MTP=3 / DFlash2 tok/s): - R9700 on iPadOS (LemonSeed Studio): 31.2 / 108.7 / 158.5 - R9700 on macOS: 32.2 / 111.9 / 159.4 - R9700 on Linux: 32.0 / 111.0 / 144.6 - Strix Halo (Radeon 8060S) on Linux: 14.0 / 47.8 / 64.4
Prefill at 4K tokens: 1,636 (iPadOS), 1,634 (macOS), 1,418 (Linux R9700), 517 (Strix Halo). It holds up at 32K: 1,395 / 1,401 / 1,226 / 462.
New in 0.5.8: - DFlash2 draft trees on by default: one target pass verifies a tree of draft candidates when that's measured to be faster than a chain - Strix Halo (gfx1151) prefill and decode work: fused gate/up GEMM, two query tiles per workgroup in prefill attention - Fixed a startup GPU fault - Linux archive bundles its own HSA runtime, so you only need the amdgpu driver with /dev/kfd access - lse-server models / pull: grab Hugging Face models with their MTP or DFlash2 companions