Loose-Info.com
Last Update 2026/02/09
TOP - 各種テスト - LLM - ローカルLLMの実測値比較 Gemma 3 (英語プロンプト)

低スペック寄りのPCでローカルLLMを動作させた際の記録です。
LLM以外の仮想マシンなどが起動され、多少負荷がかかった状態で実行しています。
ベンチマークなどでLLMの性能を評価する内容ではありません。

検証用PC

OS

Debian GNU/Linux 12 (bookworm)

CPU

Intel(R) Core(TM) i5-14400F

GPU

GeForce RTX 3060 12GB

メモリ

DDR4 PC4-25600 32GB × 4

SSD

crucial P310 CT1000P310SSD8-JP


構築環境 : Docker + Ollama (特別な設定などは無い状態)

検証用プロンプト

Could you please recommend some great places in the US to see beautiful scenery? Around 10 places in all four directions.

Gemma 3 (英語プロンプト)

GPU無し 事前のモデルのロード無し
270m(116TPS)   1b(42.9TPS)   4b(14.4TPS)   12b(5.02TPS)   27b(2.27TPS)  
GPU使用 事前のモデルのロード無し
270m(388TPS)   1b(209TPS)   4b(92.9TPS)   12b(34.7TPS)   27b(5.09TPS)  
GPU使用 事前にモデルをロード済み
270m(376TPS)   1b(207TPS)   4b(93.5TPS)   12b(35.1TPS)   27b(5.30TPS)  

TPS(tokens/s) は eval_count / eval_duration により算出

gemma3:270m(GPU無し 事前のモデルのロード無し)

Model parameters 268.10M context length 32768 embedding length 640 quantization Q8_0 2026-02-09 total_duration(合計時間) : 6236881916 (6.237s) load_duration(モデルのロード時間) : 511225010 (0.511s) prompt_eval_count(評価されたプロンプトのトークン数) : 34 prompt_eval_duration(プロンプトの評価時間) : 33512492 (0.034s) eval_count(生成トークン数) : 627 eval_duration(生成時間) : 5407044273 (5.407s) real 0m6.255s user 0m0.046s sys 0m0.005s メモリ使用量(RSS) : 607568 KB

gemma3:1b(GPU無し 事前のモデルのロード無し)

Model parameters 999.89M context length 32768 embedding length 1152 quantization Q4_K_M 2026-02-09 total_duration(合計時間) : 34119177814 (34.119s) load_duration(モデルのロード時間) : 666245455 ( 0.666s) prompt_eval_count(評価されたプロンプトのトークン数) : 34 prompt_eval_duration(プロンプトの評価時間) : 153496393 ( 0.153s) eval_count(生成トークン数) : 1388 eval_duration(生成時間) : 32325306029 (32.325s) real 0m34.130s user 0m0.028s sys 0m0.010s メモリ使用量(RSS) : 1296924 KB

gemma3:4b(GPU無し 事前のモデルのロード無し)

Model parameters 4.3B context length 131072 embedding length 2560 quantization Q4_K_M 2026-02-09 total_duration(合計時間) : 93498849101 (93.499s) load_duration(モデルのロード時間) : 1218348114 ( 1.218s) prompt_eval_count(評価されたプロンプトのトークン数) : 34 prompt_eval_duration(プロンプトの評価時間) : 550273072 ( 0.550s) eval_count(生成トークン数) : 1296 eval_duration(生成時間) : 90867468220 (90.087s) real 1m33.519s user 0m0.041s sys 0m0.022s メモリ使用量(RSS) : 4267844 KB

gemma3:12b(GPU無し 事前のモデルのロード無し)

Model parameters 12.2B context length 131072 embedding length 3840 quantization Q4_K_M 2026-02-09 total_duration(合計時間) : 261855061486 (261.855s) load_duration(モデルのロード時間) : 2061841858 ( 2.062s) prompt_eval_count(評価されたプロンプトのトークン数) : 34 prompt_eval_duration(プロンプトの評価時間) : 1807231730 ( 1.807s) eval_count(生成トークン数) : 1290 eval_duration(生成時間) : 257043274598 (257.043s) real 4m21.874s user 0m0.053s sys 0m0.031s メモリ使用量(RSS) : 9797040 KB

gemma3:27b(GPU無し 事前のモデルのロード無し)

Model parameters 27.4B context length 131072 embedding length 5376 quantization Q4_K_M 2026-02-09 total_duration(合計時間) : 700761934524 (700.762s) load_duration(モデルのロード時間) : 2843180232 ( 2.843s) prompt_eval_count(評価されたプロンプトのトークン数) : 34 prompt_eval_duration(プロンプトの評価時間) : 4215393733 ( 4.215s) eval_count(生成トークン数) : 1569 eval_duration(生成時間) : 692629375273 (692.629s) real 11m40.781s user 0m0.078s sys 0m0.067s メモリ使用量(RSS) : 19337372 KB

gemma3:270m(GPU使用 事前のモデルのロード無し)

Model parameters 268.10M context length 32768 embedding length 640 quantization Q8_0 2026-02-09 total_duration(合計時間) : 1596304003 (1.596s) load_duration(モデルのロード時間) : 633766545 (0.634s) prompt_eval_count(評価されたプロンプトのトークン数) : 34 prompt_eval_duration(プロンプトの評価時間) : 6197862 (0.006s) eval_count(生成トークン数) : 302 eval_duration(生成時間) : 777177938 (0.777s) real 0m1.607s user 0m0.026s sys 0m0.004s +---------------------------------------------------------------------------------------+ | NVIDIA-SMI 535.261.03 Driver Version: 535.261.03 CUDA Version: 12.2 | |-----------------------------------------+----------------------+----------------------+ | GPU Name Persistence-M | Bus-Id Disp.A | Volatile Uncorr. ECC | | Fan Temp Perf Pwr:Usage/Cap | Memory-Usage | GPU-Util Compute M. | | | | MIG M. | |=========================================+======================+======================| | 0 NVIDIA GeForce RTX 3060 On | 00000000:01:00.0 On | N/A | | 0% 38C P2 89W / 170W | 872MiB / 12288MiB | 76% Default | | | | N/A | +-----------------------------------------+----------------------+----------------------+ +---------------------------------------------------------------------------------------+ | Processes: | | GPU GI CI PID Type Process name GPU Memory | | ID ID Usage | |=======================================================================================| | 0 N/A N/A 1233 G /usr/lib/xorg/Xorg 102MiB | | 0 N/A N/A 1904 G xfwm4 2MiB | | 0 N/A N/A 2404 G /usr/bin/x-www-browser 244MiB | | 0 N/A N/A 43856 C /usr/bin/ollama 510MiB | +---------------------------------------------------------------------------------------+ メモリ使用量(RSS) : 622352 KB

gemma3:1b(GPU使用 事前のモデルのロード無し)

Model parameters 999.89M context length 32768 embedding length 1152 quantization Q4_K_M 2026-02-09 total_duration(合計時間) : 5917718774 (5.592s) load_duration(モデルのロード時間) : 794184855 (0.794s) prompt_eval_count(評価されたプロンプトのトークン数) : 34 prompt_eval_duration(プロンプトの評価時間) : 12585984 (0.013s) eval_count(生成トークン数) : 984 eval_duration(生成時間) : 4708182014 (4.708s) real 0m5.929s user 0m0.023s sys 0m0.008s +---------------------------------------------------------------------------------------+ | NVIDIA-SMI 535.261.03 Driver Version: 535.261.03 CUDA Version: 12.2 | |-----------------------------------------+----------------------+----------------------+ | GPU Name Persistence-M | Bus-Id Disp.A | Volatile Uncorr. ECC | | Fan Temp Perf Pwr:Usage/Cap | Memory-Usage | GPU-Util Compute M. | | | | MIG M. | |=========================================+======================+======================| | 0 NVIDIA GeForce RTX 3060 On | 00000000:01:00.0 On | N/A | | 0% 43C P2 140W / 170W | 1428MiB / 12288MiB | 87% Default | | | | N/A | +-----------------------------------------+----------------------+----------------------+ +---------------------------------------------------------------------------------------+ | Processes: | | GPU GI CI PID Type Process name GPU Memory | | ID ID Usage | |=======================================================================================| | 0 N/A N/A 1233 G /usr/lib/xorg/Xorg 102MiB | | 0 N/A N/A 1904 G xfwm4 2MiB | | 0 N/A N/A 2404 G /usr/bin/x-www-browser 240MiB | | 0 N/A N/A 43926 C /usr/bin/ollama 1070MiB | +---------------------------------------------------------------------------------------+ メモリ使用量(RSS) : 836268 KB

gemma3:4b(GPU使用 事前のモデルのロード無し)

Model parameters 4.3B context length 131072 embedding length 2560 quantization Q4_K_M 2026-02-09 total_duration(合計時間) : 15542209004 (15.542s) load_duration(モデルのロード時間) : 1319297480 ( 1.319s) prompt_eval_count(評価されたプロンプトのトークン数) : 34 prompt_eval_duration(プロンプトの評価時間) : 23731397 ( 0.024s) eval_count(生成トークン数) : 1271 eval_duration(生成時間) : 13682981582 (13.683s) real 0m15.553s user 0m0.029s sys 0m0.006s +---------------------------------------------------------------------------------------+ | NVIDIA-SMI 535.261.03 Driver Version: 535.261.03 CUDA Version: 12.2 | |-----------------------------------------+----------------------+----------------------+ | GPU Name Persistence-M | Bus-Id Disp.A | Volatile Uncorr. ECC | | Fan Temp Perf Pwr:Usage/Cap | Memory-Usage | GPU-Util Compute M. | | | | MIG M. | |=========================================+======================+======================| | 0 NVIDIA GeForce RTX 3060 On | 00000000:01:00.0 On | N/A | | 0% 52C P2 169W / 170W | 4220MiB / 12288MiB | 93% Default | | | | N/A | +-----------------------------------------+----------------------+----------------------+ +---------------------------------------------------------------------------------------+ | Processes: | | GPU GI CI PID Type Process name GPU Memory | | ID ID Usage | |=======================================================================================| | 0 N/A N/A 1233 G /usr/lib/xorg/Xorg 102MiB | | 0 N/A N/A 1904 G xfwm4 2MiB | | 0 N/A N/A 2404 G /usr/bin/x-www-browser 240MiB | | 0 N/A N/A 44042 C /usr/bin/ollama 3862MiB | +---------------------------------------------------------------------------------------+ メモリ使用量(RSS) : 1202612 KB

gemma3:12b(GPU使用 事前のモデルのロード無し)

Model parameters 12.2B context length 131072 embedding length 3840 quantization Q4_K_M 2026-02-09 total_duration(合計時間) : 58033132768 (58.033s) load_duration(モデルのロード時間) : 1866381263 ( 1.866s) prompt_eval_count(評価されたプロンプトのトークン数) : 34 prompt_eval_duration(プロンプトの評価時間) : 55041384 ( 0.055s) eval_count(生成トークン数) : 1921 eval_duration(生成時間) : 55290940407 (55.291s) real 0m58.044s user 0m0.028s sys 0m0.012s +---------------------------------------------------------------------------------------+ | NVIDIA-SMI 535.261.03 Driver Version: 535.261.03 CUDA Version: 12.2 | |-----------------------------------------+----------------------+----------------------+ | GPU Name Persistence-M | Bus-Id Disp.A | Volatile Uncorr. ECC | | Fan Temp Perf Pwr:Usage/Cap | Memory-Usage | GPU-Util Compute M. | | | | MIG M. | |=========================================+======================+======================| | 0 NVIDIA GeForce RTX 3060 On | 00000000:01:00.0 On | N/A | | 32% 60C P2 169W / 170W | 9263MiB / 12288MiB | 97% Default | | | | N/A | +-----------------------------------------+----------------------+----------------------+ +---------------------------------------------------------------------------------------+ | Processes: | | GPU GI CI PID Type Process name GPU Memory | | ID ID Usage | |=======================================================================================| | 0 N/A N/A 1233 G /usr/lib/xorg/Xorg 102MiB | | 0 N/A N/A 1904 G xfwm4 2MiB | | 0 N/A N/A 2404 G /usr/bin/x-www-browser 237MiB | | 0 N/A N/A 53781 C /usr/bin/ollama 8908MiB | +---------------------------------------------------------------------------------------+ メモリ使用量(RSS) : 1535084 KB

gemma3:27b(GPU使用 事前のモデルのロード無し)

Model parameters 27.4B context length 131072 embedding length 5376 quantization Q4_K_M 2026-02-09 total_duration(合計時間) : 269551658118 (269.552s) load_duration(モデルのロード時間) : 3336726221 ( 3.337s) prompt_eval_count(評価されたプロンプトのトークン数) : 34 prompt_eval_duration(プロンプトの評価時間) : 527759619 ( 0.528s) eval_count(生成トークン数) : 1347 eval_duration(生成時間) : 264743848374 (264.744s) real 4m29.570s user 0m0.053s sys 0m0.027s +---------------------------------------------------------------------------------------+ | NVIDIA-SMI 535.261.03 Driver Version: 535.261.03 CUDA Version: 12.2 | |-----------------------------------------+----------------------+----------------------+ | GPU Name Persistence-M | Bus-Id Disp.A | Volatile Uncorr. ECC | | Fan Temp Perf Pwr:Usage/Cap | Memory-Usage | GPU-Util Compute M. | | | | MIG M. | |=========================================+======================+======================| | 0 NVIDIA GeForce RTX 3060 On | 00000000:01:00.0 On | N/A | | 32% 54C P2 68W / 170W | 11523MiB / 12288MiB | 21% Default | | | | N/A | +-----------------------------------------+----------------------+----------------------+ +---------------------------------------------------------------------------------------+ | Processes: | | GPU GI CI PID Type Process name GPU Memory | | ID ID Usage | |=======================================================================================| | 0 N/A N/A 1233 G /usr/lib/xorg/Xorg 107MiB | | 0 N/A N/A 1904 G xfwm4 2MiB | | 0 N/A N/A 2404 G /usr/bin/x-www-browser 134MiB | | 0 N/A N/A 63278 C /usr/bin/ollama 11266MiB | +---------------------------------------------------------------------------------------+ メモリ使用量(RSS) : 8769500 KB

gemma3:270m(GPU使用 事前にモデルをロード済み)

Model parameters 268.10M context length 32768 embedding length 640 quantization Q8_0 2026-02-09 total_duration(合計時間) : 1228558870 (1.229s) load_duration(モデルのロード時間) : 67982677 (0.068s) prompt_eval_count(評価されたプロンプトのトークン数) : 34 prompt_eval_duration(プロンプトの評価時間) : 6146297 (0.006s) eval_count(生成トークン数) : 367 eval_duration(生成時間) : 977018558 (0.977s) real 0m1.239s user 0m0.023s sys 0m0.005s +---------------------------------------------------------------------------------------+ | NVIDIA-SMI 535.261.03 Driver Version: 535.261.03 CUDA Version: 12.2 | |-----------------------------------------+----------------------+----------------------+ | GPU Name Persistence-M | Bus-Id Disp.A | Volatile Uncorr. ECC | | Fan Temp Perf Pwr:Usage/Cap | Memory-Usage | GPU-Util Compute M. | | | | MIG M. | |=========================================+======================+======================| | 0 NVIDIA GeForce RTX 3060 On | 00000000:01:00.0 On | N/A | | 0% 49C P2 104W / 170W | 868MiB / 12288MiB | 77% Default | | | | N/A | +-----------------------------------------+----------------------+----------------------+ +---------------------------------------------------------------------------------------+ | Processes: | | GPU GI CI PID Type Process name GPU Memory | | ID ID Usage | |=======================================================================================| | 0 N/A N/A 1233 G /usr/lib/xorg/Xorg 102MiB | | 0 N/A N/A 1904 G xfwm4 2MiB | | 0 N/A N/A 2404 G /usr/bin/x-www-browser 240MiB | | 0 N/A N/A 44131 C /usr/bin/ollama 510MiB | +---------------------------------------------------------------------------------------+ メモリ使用量(RSS) : 622084 KB

gemma3:1b(GPU使用 事前にモデルをロード済み)

Model parameters 999.89M context length 32768 embedding length 1152 quantization Q4_K_M 2026-02-09 total_duration(合計時間) : 5802167468 (5.802s) load_duration(モデルのロード時間) : 144326848 (0.144s) prompt_eval_count(評価されたプロンプトのトークン数) : 34 prompt_eval_duration(プロンプトの評価時間) : 11807284 (0.012s) eval_count(生成トークン数) : 1080 eval_duration(生成時間) : 5212804101 (5.213s) real 0m5.813s user 0m0.028s sys 0m0.005s +---------------------------------------------------------------------------------------+ | NVIDIA-SMI 535.261.03 Driver Version: 535.261.03 CUDA Version: 12.2 | |-----------------------------------------+----------------------+----------------------+ | GPU Name Persistence-M | Bus-Id Disp.A | Volatile Uncorr. ECC | | Fan Temp Perf Pwr:Usage/Cap | Memory-Usage | GPU-Util Compute M. | | | | MIG M. | |=========================================+======================+======================| | 0 NVIDIA GeForce RTX 3060 On | 00000000:01:00.0 On | N/A | | 0% 54C P2 144W / 170W | 1428MiB / 12288MiB | 87% Default | | | | N/A | +-----------------------------------------+----------------------+----------------------+ +---------------------------------------------------------------------------------------+ | Processes: | | GPU GI CI PID Type Process name GPU Memory | | ID ID Usage | |=======================================================================================| | 0 N/A N/A 1233 G /usr/lib/xorg/Xorg 102MiB | | 0 N/A N/A 1904 G xfwm4 2MiB | | 0 N/A N/A 2404 G /usr/bin/x-www-browser 240MiB | | 0 N/A N/A 44204 C /usr/bin/ollama 1070MiB | +---------------------------------------------------------------------------------------+ メモリ使用量(RSS) : 841676 KB

gemma3:4b(GPU使用 事前にモデルをロード済み)

Model parameters 4.3B context length 131072 embedding length 2560 quantization Q4_K_M 2026-02-09 total_duration(合計時間) : 13041084068 (13.041s) load_duration(モデルのロード時間) : 141890009 ( 0.142s) prompt_eval_count(評価されたプロンプトのトークン数) : 34 prompt_eval_duration(プロンプトの評価時間) : 23355209 ( 0.023s) eval_count(生成トークン数) : 1162 eval_duration(生成時間) : 12422058113 (12.422s) real 0m13.052s user 0m0.016s sys 0m0.015s +---------------------------------------------------------------------------------------+ | NVIDIA-SMI 535.261.03 Driver Version: 535.261.03 CUDA Version: 12.2 | |-----------------------------------------+----------------------+----------------------+ | GPU Name Persistence-M | Bus-Id Disp.A | Volatile Uncorr. ECC | | Fan Temp Perf Pwr:Usage/Cap | Memory-Usage | GPU-Util Compute M. | | | | MIG M. | |=========================================+======================+======================| | 0 NVIDIA GeForce RTX 3060 On | 00000000:01:00.0 On | N/A | | 0% 61C P2 169W / 170W | 4220MiB / 12288MiB | 94% Default | | | | N/A | +-----------------------------------------+----------------------+----------------------+ +---------------------------------------------------------------------------------------+ | Processes: | | GPU GI CI PID Type Process name GPU Memory | | ID ID Usage | |=======================================================================================| | 0 N/A N/A 1233 G /usr/lib/xorg/Xorg 102MiB | | 0 N/A N/A 1904 G xfwm4 2MiB | | 0 N/A N/A 2404 G /usr/bin/x-www-browser 240MiB | | 0 N/A N/A 44280 C /usr/bin/ollama 3862MiB | +---------------------------------------------------------------------------------------+ メモリ使用量(RSS) : 1191352 KB

gemma3:12b(GPU使用 事前にモデルをロード済み)

Model parameters 12.2B context length 131072 embedding length 3840 quantization Q4_K_M 2026-02-09 total_duration(合計時間) : 40499271394 (40.499s) load_duration(モデルのロード時間) : 152698343 ( 0.153s) prompt_eval_count(評価されたプロンプトのトークン数) : 34 prompt_eval_duration(プロンプトの評価時間) : 60492057 ( 0.060s) eval_count(生成トークン数) : 1392 eval_duration(生成時間) : 39688997411 (39.689s) real 0m40.518s user 0m0.042s sys 0m0.011s +---------------------------------------------------------------------------------------+ | NVIDIA-SMI 535.261.03 Driver Version: 535.261.03 CUDA Version: 12.2 | |-----------------------------------------+----------------------+----------------------+ | GPU Name Persistence-M | Bus-Id Disp.A | Volatile Uncorr. ECC | | Fan Temp Perf Pwr:Usage/Cap | Memory-Usage | GPU-Util Compute M. | | | | MIG M. | |=========================================+======================+======================| | 0 NVIDIA GeForce RTX 3060 On | 00000000:01:00.0 On | N/A | | 32% 61C P2 169W / 170W | 9168MiB / 12288MiB | 97% Default | | | | N/A | +-----------------------------------------+----------------------+----------------------+ +---------------------------------------------------------------------------------------+ | Processes: | | GPU GI CI PID Type Process name GPU Memory | | ID ID Usage | |=======================================================================================| | 0 N/A N/A 1233 G /usr/lib/xorg/Xorg 102MiB | | 0 N/A N/A 1904 G xfwm4 2MiB | | 0 N/A N/A 2404 G /usr/bin/x-www-browser 142MiB | | 0 N/A N/A 53885 C /usr/bin/ollama 8908MiB | +---------------------------------------------------------------------------------------+ メモリ使用量(RSS) : 1564932 KB

gemma3:27b(GPU使用 事前にモデルをロード済み)

Model parameters 27.4B context length 131072 embedding length 5376 quantization Q4_K_M 2026-02-09 total_duration(合計時間) : 251644791489 (251.645s) load_duration(モデルのロード時間) : 144991637 ( 0.145s) prompt_eval_count(評価されたプロンプトのトークン数) : 34 prompt_eval_duration(プロンプトの評価時間) : 495225880 ( 0.495s) eval_count(生成トークン数) : 1325 eval_duration(生成時間) : 250158195083 (250.158s) real 4m11.656s user 0m0.035s sys 0m0.026s +---------------------------------------------------------------------------------------+ | NVIDIA-SMI 535.261.03 Driver Version: 535.261.03 CUDA Version: 12.2 | |-----------------------------------------+----------------------+----------------------+ | GPU Name Persistence-M | Bus-Id Disp.A | Volatile Uncorr. ECC | | Fan Temp Perf Pwr:Usage/Cap | Memory-Usage | GPU-Util Compute M. | | | | MIG M. | |=========================================+======================+======================| | 0 NVIDIA GeForce RTX 3060 On | 00000000:01:00.0 On | N/A | | 0% 58C P2 71W / 170W | 11785MiB / 12288MiB | 25% Default | | | | N/A | +-----------------------------------------+----------------------+----------------------+ +---------------------------------------------------------------------------------------+ | Processes: | | GPU GI CI PID Type Process name GPU Memory | | ID ID Usage | |=======================================================================================| | 0 N/A N/A 1233 G /usr/lib/xorg/Xorg 107MiB | | 0 N/A N/A 1904 G xfwm4 2MiB | | 0 N/A N/A 2404 G /usr/bin/x-www-browser 132MiB | | 0 N/A N/A 77047 C /usr/bin/ollama 11530MiB | +---------------------------------------------------------------------------------------+ メモリ使用量(RSS) : 8499584 KB