⚡
Performance-Benchmark
Wie schnell ist das Modell?
Roher Durchsatz auf echter Hardware – wie viele Token ein Modell pro Sekunde erzeugt, wie schnell es den Prompt verarbeitet und wie kurz die Zeit bis zum ersten Token ausfällt. Je höher, desto besser.
⚡ Generation (tok/s)🔄 Prefill⏱️ TTFT👥 Concurrency
| # | Modell / Hersteller | Messwerte | Parallel | GPU / CPU / RAM | Runtime | ||
|---|---|---|---|---|---|---|---|
| 1 | Qwen2.5-0.5B-Instruct0.5BQwen (Alibaba) PerformancebenchmarkTimebench 3 - Kombi (Prefill + Generation) | 256,25 tok/s TG Prefill 10.267 · TTFT 249 ms | 1× | Keine GPU (CPU-only)Apple M6 (unified CPU/GPU SoC) · 24 GB RAM | Ollamaopenclaw_cliQ4_K_M | Details → | |
| 2 | Qwen3-0.6B0.6BQwen (Alibaba) PerformancebenchmarkTimebench 3 - Kombi (Prefill + Generation) | 164,83 tok/s TG Prefill 7.014 · TTFT 390 ms | 1× | Keine GPU (CPU-only)Apple M6 (unified CPU/GPU SoC) · 24 GB RAM | Ollamaopenclaw_cliQ4_K_M | Details → | |
| 3 | gemma-4-E2B-it5BGoogle PerformancebenchmarkTimebench 3 - Kombi (Prefill + Generation) | 119,93 tok/s TG Prefill 4.370 · TTFT 507 ms | 1× | Keine GPU (CPU-only)Apple M6 (unified CPU/GPU SoC) · 24 GB RAM | Ollamaopenclaw_cliNVFP4 | Details → | |
| 4 | Qwen2.5-1.5B-Instruct1.5BQwen (Alibaba) PerformancebenchmarkTimebench 3 - Kombi (Prefill + Generation) | 113,29 tok/s TG Prefill 3.554 · TTFT 706 ms | 1× | Keine GPU (CPU-only)Apple M6 (unified CPU/GPU SoC) · 24 GB RAM | Ollamaopenclaw_cliQ4_K_M | Details → | |
| 5 | Llama-3.2-1B-Instruct1BMeta PerformancebenchmarkTimebench 3 - Kombi (Prefill + Generation) | 97,11 tok/s TG Prefill 5.083 · TTFT 500 ms | 1× | Keine GPU (CPU-only)Apple M6 (unified CPU/GPU SoC) · 24 GB RAM | Ollamaopenclaw_cliQ8_0 | Details → | |
| 6 | Qwen3-1.7B1.7BQwen (Alibaba) PerformancebenchmarkTimebench 3 - Kombi (Prefill + Generation) | 85,48 tok/s TG Prefill 3.232 · TTFT 815 ms | 1× | Keine GPU (CPU-only)Apple M6 (unified CPU/GPU SoC) · 24 GB RAM | Ollamaopenclaw_cliQ4_K_M | Details → | |
| 7 | gemma-4-E4B-it8BGoogle PerformancebenchmarkTimebench 3 - Kombi (Prefill + Generation) | 85,23 tok/s TG Prefill 2.052 · TTFT 990 ms | 1× | Keine GPU (CPU-only)Apple M6 (unified CPU/GPU SoC) · 24 GB RAM | Ollamaopenclaw_cliNVFP4 | Details → | |
| 8 | gemma-2-2b-it2BGoogle PerformancebenchmarkTimebench 3 - Kombi (Prefill + Generation) | 68,10 tok/s TG Prefill 2.491 · TTFT 768 ms | 1× | Keine GPU (CPU-only)Apple M6 (unified CPU/GPU SoC) · 24 GB RAM | Ollamaopenclaw_cliQ4_0 | Details → | |
| 9 | gemma-4-E4B-it8BGoogle PerformancebenchmarkTimebench 3 - Kombi (Prefill + Generation) | 64,65 tok/s TG Prefill 1.873 · TTFT 1.123 ms | 1× | Keine GPU (CPU-only)Apple M6 (unified CPU/GPU SoC) · 24 GB RAM | Ollamaopenclaw_cliQ4_K_M | Details → | |
| 10 | Qwen2.5-3B-Instruct3BQwen (Alibaba) PerformancebenchmarkTimebench 3 - Kombi (Prefill + Generation) | 57,50 tok/s TG Prefill 1.718 · TTFT 1.461 ms | 1× | Keine GPU (CPU-only)Apple M6 (unified CPU/GPU SoC) · 24 GB RAM | Ollamaopenclaw_cliQ4_K_M | Details → | |
| 11 | SmolLM2-1.7B-Instruct1.7BHuggingFaceTB PerformancebenchmarkTimebench 3 - Kombi (Prefill + Generation) | 54,97 tok/s TG Prefill 2.966 · TTFT 1.194 ms | 1× | Keine GPU (CPU-only)Apple M6 (unified CPU/GPU SoC) · 24 GB RAM | Ollamaopenclaw_cliQ8_0 | Details → | |
| 12 | Llama-3.2-3B-Instruct3BMeta PerformancebenchmarkTimebench 3 - Kombi (Prefill + Generation) | 52,92 tok/s TG Prefill 1.698 · TTFT 1.508 ms | 1× | Keine GPU (CPU-only)Apple M6 (unified CPU/GPU SoC) · 24 GB RAM | Ollamaopenclaw_cliQ4_K_M | Details → | |
| 13 | NVIDIA-Nemotron-3-Nano-4B4BNVIDIA PerformancebenchmarkTimebench 3 - Kombi (Prefill + Generation) | 42,29 tok/s TG Prefill 1.495 · TTFT 1.438 ms | 1× | Keine GPU (CPU-only)Apple M6 (unified CPU/GPU SoC) · 24 GB RAM | Ollamaopenclaw_cliNVFP4 | Details → | |
| 14 | Qwen3-4B4BQwen (Alibaba) PerformancebenchmarkTimebench 3 - Kombi (Prefill + Generation) | 41,82 tok/s TG Prefill 1.253 · TTFT 2.021 ms | 1× | Keine GPU (CPU-only)Apple M6 (unified CPU/GPU SoC) · 24 GB RAM | Ollamaopenclaw_cliQ4_K_M | Details → | |
| 15 | Gemma-4-26B-A4B-it26BGoogle PerformancebenchmarkTimebench 3 - Kombi (Prefill + Generation) | 41,55 tok/s TG Prefill 1.278 · TTFT 1.556 ms | 1× | Keine GPU (CPU-only)Apple M6 (unified CPU/GPU SoC) · 24 GB RAM | Ollamaopenclaw_cliNVFP4 | Details → | |
| 16 | gemma-3n-E4B-it8BGoogle PerformancebenchmarkTimebench 3 - Kombi (Prefill + Generation) | 39,36 tok/s TG Prefill 838 · TTFT 2.388 ms | 1× | Keine GPU (CPU-only)Apple M6 (unified CPU/GPU SoC) · 24 GB RAM | Ollamaopenclaw_cliQ4_K_M | Details → | |
| 17 | NVIDIA-Nemotron-3-Nano-4B4BNVIDIA PerformancebenchmarkTimebench 3 - Kombi (Prefill + Generation) | 35,84 tok/s TG Prefill 769 · TTFT 2.861 ms | 1× | Keine GPU (CPU-only)Apple M6 (unified CPU/GPU SoC) · 24 GB RAM | Ollamaopenclaw_cliQ4_K_M | Details → | |
| 18 | gpt-oss-20b20BOpenAI PerformancebenchmarkTimebench 3 - Kombi (Prefill + Generation) | 34,84 tok/s TG Prefill 1.030 · TTFT 2.057 ms | 1× | Keine GPU (CPU-only)Apple M6 (unified CPU/GPU SoC) · 24 GB RAM | Ollamaopenclaw_cliMXFP4 | Details → | |
| 19 | gemma-4-12B-it12BGoogle PerformancebenchmarkTimebench 3 - Kombi (Prefill + Generation) | 31,69 tok/s TG Prefill 616 · TTFT 3.219 ms | 1× | Keine GPU (CPU-only)Apple M6 (unified CPU/GPU SoC) · 24 GB RAM | Ollamaopenclaw_cliNVFP4 | Details → | |
| 20 | gemma-4-12B-it12BGoogle PerformancebenchmarkTimebench 3 - Kombi (Prefill + Generation) | 28,63 tok/s TG Prefill 364 · TTFT 5.626 ms | 1× | Keine GPU (CPU-only)Apple M6 (unified CPU/GPU SoC) · 24 GB RAM | Ollamaopenclaw_cliQ4_K_M | Details → | |
| 21 | Qwen2.5-7B-Instruct7BQwen (Alibaba) PerformancebenchmarkTimebench 3 - Kombi (Prefill + Generation) | 28,30 tok/s TG Prefill 637 · TTFT 3.902 ms | 1× | Keine GPU (CPU-only)Apple M6 (unified CPU/GPU SoC) · 24 GB RAM | Ollamaopenclaw_cliQ4_K_M | Details → | |
| 22 | Qwen2.5-Coder-7B-Instruct7.6BQwen (Alibaba) PerformancebenchmarkTimebench 3 - Kombi (Prefill + Generation) | 28,21 tok/s TG Prefill 722 · TTFT 3.452 ms | 1× | Keine GPU (CPU-only)Apple M6 (unified CPU/GPU SoC) · 24 GB RAM | Ollamaopenclaw_cliQ4_K_M | Details → | |
| 23 | Meta-Llama-3.1-8B-Instruct8BMeta PerformancebenchmarkTimebench 3 - Kombi (Prefill + Generation) | 26,14 tok/s TG Prefill 582 · TTFT 4.250 ms | 1× | Keine GPU (CPU-only)Apple M6 (unified CPU/GPU SoC) · 24 GB RAM | Ollamaopenclaw_cliQ4_K_M | Details → | |
| 24 | Qwen3-8B8.2BQwen (Alibaba) PerformancebenchmarkTimebench 3 - Kombi (Prefill + Generation) | 24,96 tok/s TG Prefill 745 · TTFT 3.467 ms | 1× | Keine GPU (CPU-only)Apple M6 (unified CPU/GPU SoC) · 24 GB RAM | Ollamaopenclaw_cliQ4_K_M | Details → | |
| 25 | gemma-2-9b-it9BGoogle PerformancebenchmarkTimebench 3 - Kombi (Prefill + Generation) | 20,31 tok/s TG Prefill 550 · TTFT 3.503 ms | 1× | Keine GPU (CPU-only)Apple M6 (unified CPU/GPU SoC) · 24 GB RAM | Ollamaopenclaw_cliQ4_0 | Details → | |
| 26 | Qwen3-14B14BQwen (Alibaba) PerformancebenchmarkTimebench 3 - Kombi (Prefill + Generation) | 14,03 tok/s TG Prefill 321 · TTFT 8.097 ms | 1× | Keine GPU (CPU-only)Apple M6 (unified CPU/GPU SoC) · 24 GB RAM | Ollamaopenclaw_cliQ4_K_M | Details → | |
| 27 | Codestral-22B-v0.122BMistral AI PerformancebenchmarkTimebench 3 - Kombi (Prefill + Generation) | 8,89 tok/s TG Prefill 241 · TTFT 11.388 ms | 1× | Keine GPU (CPU-only)Apple M6 (unified CPU/GPU SoC) · 24 GB RAM | Ollamaopenclaw_cliQ4_0 | Details → | |
| 28 | Devstral-Small-2-24B-Instruct-251224BMistral AI PerformancebenchmarkTimebench 3 - Kombi (Prefill + Generation) | 8,85 tok/s TG Prefill 276 · TTFT 9.964 ms | 1× | Keine GPU (CPU-only)Apple M6 (unified CPU/GPU SoC) · 24 GB RAM | Ollamaopenclaw_cliQ4_K_M | Details → | |
| 29 | Mistral-Small-3.1-24B-Instruct-250324BMistral AI PerformancebenchmarkTimebench 3 - Kombi (Prefill + Generation) | 8,73 tok/s TG Prefill 245 · TTFT 10.366 ms | 1× | Keine GPU (CPU-only)Apple M6 (unified CPU/GPU SoC) · 24 GB RAM | Ollamaopenclaw_cliQ4_K_M | Details → | |
| 30 | Mistral-Small-24B-Instruct-250123.6BMistral AI PerformancebenchmarkTimebench 3 - Kombi (Prefill + Generation) | 8,47 tok/s TG Prefill 226 · TTFT 10.329 ms | 1× | Keine GPU (CPU-only)Apple M6 (unified CPU/GPU SoC) · 24 GB RAM | Ollamaopenclaw_cliQ4_K_M | Details → |