llama.cpp CUDA Jetson AGX Orin llama-2-7b Q4_0 tg128 FA
35.88 tok/s (tg128)
Fields
id
benchmarks/llamacpp-scoreboard-jetson-agx-orin-64gb-fa-tg128Metric
decode_tok_s
Value
35.88
Unit
tok/s (tg128)
Model
llama-2-7b.Q4_0.gguf (ggml-org, 3.56 GiB, 6.74B params)
Methodology
COMMUNITY-SCOREBOARD (llama.cpp discussion #15013, maintainer-curated CUDA backend). llama-bench llama-2-7b.Q4_0.gguf (ggml-org, 3.56 GiB, 6.74B params), pp512 and tg128, n_gpu_layers 99, batch 1, full GPU offload, CUDA. Per-request batch-1, not an aggregate across concurrent requests. Flash attention on (-fa 1). Build c1b1876, reporter @TinyServal. Printed mean 35.88. Re-read of the discussion page on 2026-10-17.
Measured by
third_party
Reproducible
yes
notes
Jetson AGX Orin, 64 GB / LPDDR5 / 256 bit, build c1b1876, reporter @TinyServal. token generation cell.
Notes
Jetson AGX Orin, 64 GB / LPDDR5 / 256 bit, build c1b1876, reporter @TinyServal. token generation cell.