odroid@odroid:~/llama.cpp$ pwd
/home/odroid/llama.cpp
odroid@odroid:~/llama.cpp$ wget https://huggingface.co/Qwen/Qwen2.5-0.5B-Instruct-GGUF/resolve/main/qwen2.5-0.5b-instruct-q4_k_m.gguf -O qwen0.5b.gguf
--2026-08-27 01:55:15--  https://huggingface.co/Qwen/Qwen2.5-0.5B-Instruct-GGUF/resolve/main/qwen2.5-0.5b-instruct-q4_k_m.gguf
Resolving huggingface.co (huggingface.co)... 2600:9000:28bc:c600:17:b174:6d00:93a1, 2600:9000:28bc:e200:17:b174:6d00:93a1, 2600:9000:28bc:7200:17:b174:6d00:93a1, ...
Connecting to huggingface.co (huggingface.co)|2600:9000:28bc:c600:17:b174:6d00:93a1|:443... connected.
HTTP request sent, awaiting response... 302 Found
Location: https://us.aws.cdn.hf.co/xet-bridge-us/66e98ad5d89bf05ecbbdfe9e/0663b826e09042345d6e2069d9655cd3ae84812d8f39a5894bed7b24fc0e86cc?response-content-disposition=inline%3B+filename*%3DUTF-8%27%27qwen2.5-0.5b-instruct-q4_k_m.gguf%3B+filename%3D%22qwen2.5-0.5b-instruct-q4_k_m.gguf%22%3B&user_id=public&X-Xet-Cas-Uid=public&Expires=1787799315&Policy=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__&Signature=MEUCIFpYpgeYBZe5QlcxE510QT8x%7EW3CvarUjCdqKhbzgBmEAiEAr0SOyOm63uXWL37-bH61jEZh2UFUFkEKrAZ7FHNfadg_&Key-Pair-Id=01KXEF4KZ1B6FV465MAWR4M21F [following]
--2026-08-27 01:55:15--  https://us.aws.cdn.hf.co/xet-bridge-us/66e98ad5d89bf05ecbbdfe9e/0663b826e09042345d6e2069d9655cd3ae84812d8f39a5894bed7b24fc0e86cc?response-content-disposition=inline%3B+filename*%3DUTF-8%27%27qwen2.5-0.5b-instruct-q4_k_m.gguf%3B+filename%3D%22qwen2.5-0.5b-instruct-q4_k_m.gguf%22%3B&user_id=public&X-Xet-Cas-Uid=public&Expires=1787799315&Policy=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__&Signature=MEUCIFpYpgeYBZe5QlcxE510QT8x%7EW3CvarUjCdqKhbzgBmEAiEAr0SOyOm63uXWL37-bH61jEZh2UFUFkEKrAZ7FHNfadg_&Key-Pair-Id=01KXEF4KZ1B6FV465MAWR4M21F
Resolving us.aws.cdn.hf.co (us.aws.cdn.hf.co)... 100.29.213.216, 100.50.185.240, 100.57.83.12, ...
Connecting to us.aws.cdn.hf.co (us.aws.cdn.hf.co)|100.29.213.216|:443... connected.
HTTP request sent, awaiting response... 200 OK
Length: 491400032 (469M) [application/octet-stream]
Saving to: ‘qwen0.5b.gguf’

qwen0.5b.gguf                                     100%[============================================================================================================>] 468.63M  15.1MB/s    in 50s     

2026-08-27 01:56:05 (9.32 MB/s) - ‘qwen0.5b.gguf’ saved [491400032/491400032]

odroid@odroid:~/llama.cpp$ 
odroid@odroid:~/llama.cpp$ ./build/bin/llama-bench -m qwen0.5b.gguf -t 4
| model                          |       size |     params | backend    | threads |            test |                  t/s |
| ------------------------------ | ---------: | ---------: | ---------- | ------: | --------------: | -------------------: |
^C
odroid@odroid:~/llama.cpp$ 
odroid@odroid:~/llama.cpp$ ./build/bin/llama-bench -m qwen0.5b.gguf -t 4 --progress -r 1
| model                          |       size |     params | backend    | threads |            test |                  t/s |
| ------------------------------ | ---------: | ---------: | ---------- | ------: | --------------: | -------------------: |
llama-bench: benchmark 1/2: starting
llama-bench: benchmark 1/2: warmup prompt run
llama-bench: benchmark 1/2: prompt run 1/1
| qwen2 1B Q4_K - Medium         | 462.96 MiB |   630.17 M | CPU        |       4 |           pp512 |          4.48 ± 0.00 |
llama-bench: benchmark 2/2: starting
llama-bench: benchmark 2/2: warmup generation run
llama-bench: benchmark 2/2: generation run 1/1
| qwen2 1B Q4_K - Medium         | 462.96 MiB |   630.17 M | CPU        |       4 |           tg128 |          2.42 ± 0.00 |

build: 925e11799 (10642)
odroid@odroid:~/llama.cpp$ ./build/bin/llama-bench -m qwen0.5b.gguf -t 4 --progress -r 1
| model                          |       size |     params | backend    | threads |            test |                  t/s |
| ------------------------------ | ---------: | ---------: | ---------- | ------: | --------------: | -------------------: |
llama-bench: benchmark 1/2: starting
llama-bench: benchmark 1/2: warmup prompt run
llama-bench: benchmark 1/2: prompt run 1/1
| qwen2 1B Q4_K - Medium         | 462.96 MiB |   630.17 M | CPU        |       4 |           pp512 |          4.46 ± 0.00 |
llama-bench: benchmark 2/2: starting
llama-bench: benchmark 2/2: warmup generation run
llama-bench: benchmark 2/2: generation run 1/1

| qwen2 1B Q4_K - Medium         | 462.96 MiB |   630.17 M | CPU        |       4 |           tg128 |          2.56 ± 0.00 |

build: 925e11799 (10642)
odroid@odroid:~/llama.cpp$ 
odroid@odroid:~/llama.cpp$ ./build/bin/llama-bench -m qwen0.5b.gguf -t 4 --progress -r 3
| model                          |       size |     params | backend    | threads |            test |                  t/s |
| ------------------------------ | ---------: | ---------: | ---------- | ------: | --------------: | -------------------: |
llama-bench: benchmark 1/2: starting
llama-bench: benchmark 1/2: warmup prompt run


















odroid@odroid:~/llama.cpp$ ./build/bin/llama-bench -m qwen0.5b.gguf -t 4
| model                          |       size |     params | backend    | threads |            test |                  t/s |
| ------------------------------ | ---------: | ---------: | ---------- | ------: | --------------: | -------------------: |
^C
odroid@odroid:~/llama.cpp$ 
odroid@odroid:~/llama.cpp$ ./build/bin/llama-bench -m qwen0.5b.gguf -t 4 --progress -r 1
| model                          |       size |     params | backend    | threads |            test |                  t/s |
| ------------------------------ | ---------: | ---------: | ---------- | ------: | --------------: | -------------------: |
llama-bench: benchmark 1/2: starting
llama-bench: benchmark 1/2: warmup prompt run
llama-bench: benchmark 1/2: prompt run 1/1
| qwen2 1B Q4_K - Medium         | 462.96 MiB |   630.17 M | CPU        |       4 |           pp512 |          4.48 ± 0.00 |
llama-bench: benchmark 2/2: starting
llama-bench: benchmark 2/2: warmup generation run
llama-bench: benchmark 2/2: generation run 1/1
| qwen2 1B Q4_K - Medium         | 462.96 MiB |   630.17 M | CPU        |       4 |           tg128 |          2.42 ± 0.00 |

build: 925e11799 (10642)
odroid@odroid:~/llama.cpp$ ./build/bin/llama-bench -m qwen0.5b.gguf -t 4 --progress -r 1
| model                          |       size |     params | backend    | threads |            test |                  t/s |
| ------------------------------ | ---------: | ---------: | ---------- | ------: | --------------: | -------------------: |
llama-bench: benchmark 1/2: starting
llama-bench: benchmark 1/2: warmup prompt run
llama-bench: benchmark 1/2: prompt run 1/1
| qwen2 1B Q4_K - Medium         | 462.96 MiB |   630.17 M | CPU        |       4 |           pp512 |          4.46 ± 0.00 |
llama-bench: benchmark 2/2: starting
llama-bench: benchmark 2/2: warmup generation run
llama-bench: benchmark 2/2: generation run 1/1

| qwen2 1B Q4_K - Medium         | 462.96 MiB |   630.17 M | CPU        |       4 |           tg128 |          2.56 ± 0.00 |

build: 925e11799 (10642)
odroid@odroid:~/llama.cpp$ 
odroid@odroid:~/llama.cpp$ ./build/bin/llama-bench -m qwen0.5b.gguf -t 4 --progress -r 3
| model                          |       size |     params | backend    | threads |            test |                  t/s |
| ------------------------------ | ---------: | ---------: | ---------- | ------: | --------------: | -------------------: |
llama-bench: benchmark 1/2: starting
llama-bench: benchmark 1/2: warmup prompt run
llama-bench: benchmark 1/2: prompt run 1/3
llama-bench: benchmark 1/2: prompt run 2/3
llama-bench: benchmark 1/2: prompt run 3/3
| qwen2 1B Q4_K - Medium         | 462.96 MiB |   630.17 M | CPU        |       4 |           pp512 |          4.49 ± 0.01 |
llama-bench: benchmark 2/2: starting
llama-bench: benchmark 2/2: warmup generation run
llama-bench: benchmark 2/2: generation run 1/3
llama-bench: benchmark 2/2: generation run 2/3
llama-bench: benchmark 2/2: generation run 3/3
| qwen2 1B Q4_K - Medium         | 462.96 MiB |   630.17 M | CPU        |       4 |           tg128 |          2.58 ± 0.00 |

build: 925e11799 (10642)



odroid@odroid:~/llama.cpp$ taskset -c 4-7 ./build/bin/llama-bench -m qwen0.5b.gguf -t 4 --progress -r 1
| model                          |       size |     params | backend    | threads |            test |                  t/s |
| ------------------------------ | ---------: | ---------: | ---------- | ------: | --------------: | -------------------: |
llama-bench: benchmark 1/2: starting
llama-bench: benchmark 1/2: warmup prompt run