🏆
Ranglisten
LLM Leaderboard
Reale Messergebnisse aller getesteten Sprachmodelle auf echter, benannter Hardware – bewertet nach roher Geschwindigkeit (Performance in Token pro Sekunde, Prefill und Zeit bis zum ersten Token) und nach praktischer Aufgaben-Qualität in vollständigen Agenten- und Chat-Abläufen (Harness). Wähle unten einen Benchmark-Typ oder filtere nach Modell, Hersteller und Hardware, um genau zu sehen, was jeweils geprüft wird und wie die Ergebnisse zustande kommen.
⚡ Performance (tok/s)🤖 Harness-Qualität👥 Concurrency🖥️ echte Hardware
| # | Modell / Hersteller | Messwerte | Parallel | GPU / CPU / RAM | Runtime | ||
|---|---|---|---|---|---|---|---|
| 1 | Qwen2.5-0.5B-InstructQwen (Alibaba) PerformancebenchmarkTimebench 3 - Kombi (Prefill + Generation) | 19,64 tok/s TG Prefill 135 · TTFT 103.828 ms | 5× | Keine GPU (CPU-only) i4 x Cortex-A78C + 4 x Cortex-A55 · 15 GB RAM | llama.cppopenclaw_cliQ4_K_M | Details → | |
| 2 | Qwen2.5-0.5B-InstructQwen (Alibaba) PerformancebenchmarkTimebench 3 - Kombi (Prefill + Generation) | 19,33 tok/s TG Prefill 171 · TTFT 178.576 ms | 10× | Keine GPU (CPU-only) i4 x Cortex-A78C + 4 x Cortex-A55 · 15 GB RAM | llama.cppopenclaw_cliQ4_K_M | Details → | |
| 3 | Llama-3.2-1B-Instruct1BMeta PerformancebenchmarkTimebench 3 - Kombi (Prefill + Generation) | 13,17 tok/s TG Prefill 122 · TTFT 228.054 ms | 10× | Keine GPU (CPU-only) i4 x Cortex-A78C + 4 x Cortex-A55 · 15 GB RAM | llama.cppopenclaw_cliQ4_K_M | Details → | |
| 4 | Llama-3.2-1B-Instruct1BMeta PerformancebenchmarkTimebench 3 - Kombi (Prefill + Generation) | 12,93 tok/s TG Prefill 95 · TTFT 136.105 ms | 5× | Keine GPU (CPU-only) i4 x Cortex-A78C + 4 x Cortex-A55 · 15 GB RAM | llama.cppopenclaw_cliQ4_K_M | Details → | |
| 5 | Qwen2.5-0.5B-InstructQwen (Alibaba) PerformancebenchmarkTimebench 3 - Kombi (Prefill + Generation) | 11,74 tok/s TG Prefill 85 · TTFT 29.146 ms | 1× | Keine GPU (CPU-only) i4 x Cortex-A78C + 4 x Cortex-A55 · 15 GB RAM | llama.cppopenclaw_cliQ4_K_M | Details → | |
| 6 | Qwen2.5-1.5B-InstructQwen (Alibaba) PerformancebenchmarkTimebench 3 - Kombi (Prefill + Generation) | 9,14 tok/s TG Prefill 67 · TTFT 194.902 ms | 5× | Keine GPU (CPU-only) i4 x Cortex-A78C + 4 x Cortex-A55 · 15 GB RAM | llama.cppopenclaw_cliQ4_K_M | Details → | |
| 7 | Qwen3-0.6B0.6BQwen (Alibaba) PerformancebenchmarkTimebench 3 - Kombi (Prefill + Generation) | 8,04 tok/s TG Prefill 108 · TTFT 227.805 ms | 10× | Keine GPU (CPU-only) i4 x Cortex-A78C + 4 x Cortex-A55 · 15 GB RAM | llama.cppopenclaw_cliQ4_K_M | Details → | |
| 8 | gemma-2-2b-it2BGoogle PerformancebenchmarkTimebench 3 - Kombi (Prefill + Generation) | 7,94 tok/s TG Prefill 52 · TTFT 404.593 ms | 10× | Keine GPU (CPU-only) i4 x Cortex-A78C + 4 x Cortex-A55 · 15 GB RAM | llama.cppopenclaw_cliQ4_K_M | Details → | |
| 9 | gemma-2-2b-it2BGoogle PerformancebenchmarkTimebench 3 - Kombi (Prefill + Generation) | 7,49 tok/s TG Prefill 35 · TTFT 283.919 ms | 5× | Keine GPU (CPU-only) i4 x Cortex-A78C + 4 x Cortex-A55 · 15 GB RAM | llama.cppopenclaw_cliQ4_K_M | Details → | |
| 10 | Qwen3-0.6B0.6BQwen (Alibaba) PerformancebenchmarkTimebench 3 - Kombi (Prefill + Generation) | 7,43 tok/s TG Prefill 83 · TTFT 139.346 ms | 5× | Keine GPU (CPU-only) i4 x Cortex-A78C + 4 x Cortex-A55 · 15 GB RAM | llama.cppopenclaw_cliQ4_K_M | Details → | |
| 11 | Llama-3.2-1B-Instruct1BMeta PerformancebenchmarkTimebench 3 - Kombi (Prefill + Generation) | 7,33 tok/s TG Prefill 65 · TTFT 38.290 ms | 1× | Keine GPU (CPU-only) i4 x Cortex-A78C + 4 x Cortex-A55 · 15 GB RAM | llama.cppopenclaw_cliQ4_K_M | Details → | |
| 12 | Qwen2.5-1.5B-InstructQwen (Alibaba) PerformancebenchmarkTimebench 3 - Kombi (Prefill + Generation) | 4,93 tok/s TG Prefill 47 · TTFT 52.732 ms | 1× | Keine GPU (CPU-only) i4 x Cortex-A78C + 4 x Cortex-A55 · 15 GB RAM | llama.cppopenclaw_cliQ4_K_M | Details → | |
| 13 | Qwen2.5-1.5B-InstructQwen (Alibaba) PerformancebenchmarkTimebench 3 - Kombi (Prefill + Generation) | 4,11 tok/s TG Prefill 91 · TTFT 301.409 ms | 10× | Keine GPU (CPU-only) i4 x Cortex-A78C + 4 x Cortex-A55 · 15 GB RAM | llama.cppopenclaw_cliQ4_K_M | Details → | |
| 14 | Qwen3-1.7B1.7BQwen (Alibaba) PerformancebenchmarkTimebench 3 - Kombi (Prefill + Generation) | 4,04 tok/s TG Prefill 48 · TTFT 383.019 ms | 10× | Keine GPU (CPU-only) i4 x Cortex-A78C + 4 x Cortex-A55 · 15 GB RAM | llama.cppopenclaw_cliQ4_K_M | Details → | |
| 15 | Qwen3-1.7B1.7BQwen (Alibaba) PerformancebenchmarkTimebench 3 - Kombi (Prefill + Generation) | 3,82 tok/s TG Prefill 39 · TTFT 274.809 ms | 5× | Keine GPU (CPU-only) i4 x Cortex-A78C + 4 x Cortex-A55 · 15 GB RAM | llama.cppopenclaw_cliQ4_K_M | Details → | |
| 16 | gemma-2-2b-it2BGoogle PerformancebenchmarkTimebench 3 - Kombi (Prefill + Generation) | 3,76 tok/s TG Prefill 31 · TTFT 59.057 ms | 1× | Keine GPU (CPU-only) i4 x Cortex-A78C + 4 x Cortex-A55 · 15 GB RAM | llama.cppopenclaw_cliQ4_K_M | Details → | |
| 17 | Qwen2.5-3B-Instruct3BQwen (Alibaba) PerformancebenchmarkTimebench 3 - Kombi (Prefill + Generation) | 3,47 tok/s TG Prefill 33 · TTFT 369.084 ms | 5× | Keine GPU (CPU-only) i4 x Cortex-A78C + 4 x Cortex-A55 · 15 GB RAM | llama.cppopenclaw_cliQ4_K_M | Details → | |
| 18 | Qwen2.5-3B-Instruct3BQwen (Alibaba) PerformancebenchmarkTimebench 3 - Kombi (Prefill + Generation) | 3,22 tok/s TG Prefill 24 · TTFT 103.262 ms | 1× | Keine GPU (CPU-only) i4 x Cortex-A78C + 4 x Cortex-A55 · 15 GB RAM | llama.cppopenclaw_cliQ4_K_M | Details → | |
| 19 | Qwen3-0.6B0.6BQwen (Alibaba) PerformancebenchmarkTimebench 3 - Kombi (Prefill + Generation) | 3,17 tok/s TG Prefill 80 · TTFT 30.662 ms | 1× | Keine GPU (CPU-only) i4 x Cortex-A78C + 4 x Cortex-A55 · 15 GB RAM | llama.cppopenclaw_cliQ4_K_M | Details → | |
| 20 | SmolLM2-1.7B-Instruct1.7BHuggingFaceTB PerformancebenchmarkTimebench 3 - Kombi (Prefill + Generation) | 3,10 tok/s TG Prefill 45 · TTFT 281.637 ms | 5× | Keine GPU (CPU-only) i4 x Cortex-A78C + 4 x Cortex-A55 · 15 GB RAM | llama.cppopenclaw_cliQ4_K_M | Details → | |
| 21 | Qwen3-1.7B1.7BQwen (Alibaba) PerformancebenchmarkTimebench 3 - Kombi (Prefill + Generation) | 2,95 tok/s TG Prefill 41 · TTFT 60.139 ms | 1× | Keine GPU (CPU-only) i4 x Cortex-A78C + 4 x Cortex-A55 · 15 GB RAM | llama.cppopenclaw_cliQ4_K_M | Details → | |
| 22 | SmolLM2-1.7B-Instruct1.7BHuggingFaceTB PerformancebenchmarkTimebench 3 - Kombi (Prefill + Generation) | 2,89 tok/s TG Prefill 38 · TTFT 81.741 ms | 1× | Keine GPU (CPU-only) i4 x Cortex-A78C + 4 x Cortex-A55 · 15 GB RAM | llama.cppopenclaw_cliQ4_K_M | Details → | |
| 23 | Llama-3.2-3B-Instruct3BMeta PerformancebenchmarkTimebench 3 - Kombi (Prefill + Generation) | 2,46 tok/s TG Prefill 35 · TTFT 360.878 ms | 5× | Keine GPU (CPU-only) i4 x Cortex-A78C + 4 x Cortex-A55 · 15 GB RAM | llama.cppopenclaw_cliQ4_K_M | Details → | |
| 24 | Llama-3.2-3B-Instruct3BMeta PerformancebenchmarkTimebench 3 - Kombi (Prefill + Generation) | 1,99 tok/s TG Prefill 22 · TTFT 104.643 ms | 1× | Keine GPU (CPU-only) i4 x Cortex-A78C + 4 x Cortex-A55 · 15 GB RAM | llama.cppopenclaw_cliQ4_K_M | Details → | |
| 25 | Qwen3-8B8.2BQwen (Alibaba) PerformancebenchmarkTimebench 3 - Kombi (Prefill + Generation) | 1,62 tok/s TG Prefill 9 · TTFT 232.748 ms | 1× | Keine GPU (CPU-only) i4 x Cortex-A78C + 4 x Cortex-A55 · 15 GB RAM | llama.cppopenclaw_cliQ4_K_M | Details → | |
| 26 | Meta-Llama-3.1-8B-Instruct8BMeta PerformancebenchmarkTimebench 3 - Kombi (Prefill + Generation) | 1,19 tok/s TG Prefill 9 · TTFT 227.569 ms | 1× | Keine GPU (CPU-only) i4 x Cortex-A78C + 4 x Cortex-A55 · 15 GB RAM | llama.cppopenclaw_cliQ4_K_M | Details → | |
| 27 | Qwen3-4B4BQwen (Alibaba) PerformancebenchmarkTimebench 3 - Kombi (Prefill + Generation) | 0,91 tok/s TG Prefill 15 · TTFT 142.282 ms | 1× | Keine GPU (CPU-only) i4 x Cortex-A78C + 4 x Cortex-A55 · 15 GB RAM | llama.cppopenclaw_cliQ4_K_M | Details → | |
| 28 | Llama-3.2-3B-Instruct3BMeta PerformancebenchmarkTimebench 3 - Kombi (Prefill + Generation) | 0,09 tok/s TG Prefill 31 · TTFT 347.304 ms | 10× | Keine GPU (CPU-only) i4 x Cortex-A78C + 4 x Cortex-A55 · 15 GB RAM | llama.cppopenclaw_cliQ4_K_M | Details → | |
| 29 | Qwen2.5-3B-Instruct3BQwen (Alibaba) PerformancebenchmarkTimebench 3 - Kombi (Prefill + Generation) | 0,07 tok/s TG Prefill 33 · TTFT 360.019 ms | 10× | Keine GPU (CPU-only) i4 x Cortex-A78C + 4 x Cortex-A55 · 15 GB RAM | llama.cppopenclaw_cliQ4_K_M | Details → | |
| 30 | Meta-Llama-3.1-8B-Instruct8BMeta PerformancebenchmarkTimebench 3 - Kombi (Prefill + Generation) | 0,02 tok/s TG Prefill 7 · TTFT 453.619 ms | 5× | Keine GPU (CPU-only) i4 x Cortex-A78C + 4 x Cortex-A55 · 15 GB RAM | llama.cppopenclaw_cliQ4_K_M | Details → | |
| 31 | Qwen3-4B4BQwen (Alibaba) PerformancebenchmarkTimebench 3 - Kombi (Prefill + Generation) | 0,01 tok/s TG Prefill 6 · TTFT 506.953 ms | 5× | Keine GPU (CPU-only) i4 x Cortex-A78C + 4 x Cortex-A55 · 15 GB RAM | llama.cppopenclaw_cliQ4_K_M | Details → |