← Leaderboard LLM run · 580f60ff-097e-4f00-b4fc-7943964e3731

Qwen3.8-27B-GGUF-UD-Q4_K_XL

AMD Radeon RX 7900 XTX (gfx1100) · profile Radeon-RX-7900-XTX-WSL

LLMUD-Q4_K_XLdensedgpullamacpp-rocmvalid 2026-08-24 20:21 UTC
17.1
Decode tok/s ↑
± 0.846 (95% CI) · sd 0.682
137
Prefill tok/s ↑ (cache-assisted)
± 46.5 (95% CI) · sd 37.5
577
TTFT ms ↓ (cache-assisted)
± 218 (95% CI)
58.5
Inter-token ms ↓
4.5
Peak memory GB
5
Trials

Validity: valid

The checks LemonMatrix ran while measuring; an invalid run is shown but never ranked.

Trials

Each measured trial after warm-up, from the run's trial sidecar. The summary above is computed from these.

#ttft msdecode tokens per secprompt tokensoutput tokenshost memory gbdecode tokens per sec
1453.416.3791502.7
2439.517.26792563.2
3853.216.49791613.6
4649.617.86791624.1
5489.717.6791564.5

Same model, other runs

Every run of Qwen3.8-27B-GGUF-UD-Q4_K_XL · UD-Q4_K_XL on the board, ranked by decode tok/s. This run is highlighted.

ProfileDeviceEngineBackend decode tok/s ↑Valid
AMD-Instinct-MI350Xgfx950 GPUdgpullamacpp-rocm100.01
valid
Radeon-RX-7900-XTXAMD Radeon RX 7900 XTXdgpullamacpp-vulkan20.48
valid
Radeon-RX-7900-XTX-WSLAMD Radeon RX 7900 XTX (gfx1100)dgpullamacpp-rocm17.10
valid
Radeon-RX-7900-XTXAMD Radeon RX 7900 XTXdgpullamacpp-rocm14.19
valid
Radeon-RX-7900-XTXAMD Radeon RX 7900 XTXdgpullamacpp-vulkan9.83
invalid
NVIDIA-L4Intel(R) Xeon(R) Silver 4314 CPU @ 2.40GHzcpullamacpp-cpu3.65
valid
Radeon-RX-7900-XTXAMD Radeon RX 7900 XTXdgpullamacpp-rocm2.50
invalid

Model and configuration

model.nameQwen3.8-27B-GGUF-UD-Q4_K_XL
model.classdense
model.quantizationUD-Q4_K_XL
model.context_length4096
config.compute_enginedgpu
config.backendllamacpp-rocm
config.oslinux
config.power_stateplugged

Environment

cpuAMD Ryzen 7 5800X3D 8-Core Processor
device_modelAMD Ryzen 7 5800X3D 8-Core Processor
dgpuAMD Radeon RX 7900 XTX (gfx1100)
dgpu_familygfx1100
driver_versionunknown
lemonade_version11.6.0
memory_gb15.57
os_versionLinux-6.18.33.2-microsoft-standard-WSL2 (Ubuntu 24.04)
backend_versionb10397

All metrics

prefill.tokens_per_sec136.9
prefill.stddev37.48
prefill.p95178.6
prefill.ci95_half_width46.53
decode.tokens_per_sec17.1
decode.inter_token_latency_ms58.48
decode.stddev0.6817
decode.p9517.81
decode.ci95_half_width0.8463
ttft_ms577.1
trial_count5
peak_memory_gb4.5
prompt_sha2561c43b07d6b0d7624785966ffe2ddc076aa76b5fcf9357882ac1165f0073f43ee
ttft_ms_stddev175.6
ttft_ms_ci95_half_width218