⚡
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) | 298,13 tok/s TG Prefill 14.858 · TTFT 175 ms | 1× | Keine GPU (CPU-only)Apple M6 (unified CPU/GPU SoC, ARM64, AdvSIMD) · 24 GB RAM | lmstudio-mlxlmstudioMLX 4bit | Details → | |
| 2 | Qwen3-0.6B0.6BQwen (Alibaba) PerformancebenchmarkTimebench 3 - Kombi (Prefill + Generation) | 189,50 tok/s TG Prefill 11.438 · TTFT 251 ms | 1× | Keine GPU (CPU-only)Apple M6 (unified CPU/GPU SoC, ARM64, AdvSIMD) · 24 GB RAM | lmstudio-mlxlmstudioMLX 4bit | Details → | |
| 3 | Llama-3.2-1B-Instruct1BMeta PerformancebenchmarkTimebench 3 - Kombi (Prefill + Generation) | 133,16 tok/s TG Prefill 7.177 · TTFT 377 ms | 1× | Keine GPU (CPU-only)Apple M6 (unified CPU/GPU SoC, ARM64, AdvSIMD) · 24 GB RAM | lmstudio-mlxlmstudioMLX 4bit | Details → | |
| 4 | Qwen2.5-1.5B-Instruct1.5BQwen (Alibaba) PerformancebenchmarkTimebench 3 - Kombi (Prefill + Generation) | 126,71 tok/s TG Prefill 5.137 · TTFT 507 ms | 1× | Keine GPU (CPU-only)Apple M6 (unified CPU/GPU SoC, ARM64, AdvSIMD) · 24 GB RAM | lmstudio-mlxlmstudioMLX 4bit | Details → | |
| 5 | Qwen3-1.7B1.7BQwen (Alibaba) PerformancebenchmarkTimebench 3 - Kombi (Prefill + Generation) | 100,53 tok/s TG Prefill 4.497 · TTFT 604 ms | 1× | Keine GPU (CPU-only)Apple M6 (unified CPU/GPU SoC, ARM64, AdvSIMD) · 24 GB RAM | lmstudio-mlxlmstudioMLX 4bit | Details → | |
| 6 | gemma-2-2b-it2BGoogle PerformancebenchmarkTimebench 3 - Kombi (Prefill + Generation) | 75,08 tok/s TG Prefill 2.212 · TTFT 876 ms | 1× | Keine GPU (CPU-only)Apple M6 (unified CPU/GPU SoC, ARM64, AdvSIMD) · 24 GB RAM | lmstudio-mlxlmstudioMLX 4bit | Details → | |
| 7 | Qwen2.5-3B-Instruct3BQwen (Alibaba) PerformancebenchmarkTimebench 3 - Kombi (Prefill + Generation) | 65,98 tok/s TG Prefill 2.669 · TTFT 988 ms | 1× | Keine GPU (CPU-only)Apple M6 (unified CPU/GPU SoC, ARM64, AdvSIMD) · 24 GB RAM | lmstudio-mlxlmstudioMLX 4bit | Details → | |
| 8 | Llama-3.2-3B-Instruct3BMeta PerformancebenchmarkTimebench 3 - Kombi (Prefill + Generation) | 62,86 tok/s TG Prefill 2.562 · TTFT 1.026 ms | 1× | Keine GPU (CPU-only)Apple M6 (unified CPU/GPU SoC, ARM64, AdvSIMD) · 24 GB RAM | lmstudio-mlxlmstudioMLX 4bit | Details → | |
| 9 | NVIDIA-Nemotron-3-Nano-4B4BNVIDIA PerformancebenchmarkTimebench 3 - Kombi (Prefill + Generation) | 56,16 tok/s TG Prefill 1.002 · TTFT 2.226 ms | 1× | Keine GPU (CPU-only)Apple M6 (unified CPU/GPU SoC, ARM64, AdvSIMD) · 24 GB RAM | lmstudio-mlxlmstudioMLX 4bit | Details → | |
| 10 | gpt-oss-20b20BOpenAI PerformancebenchmarkTimebench 3 - Kombi (Prefill + Generation) | 53,05 tok/s TG Prefill 1.769 · TTFT 1.368 ms | 1× | Keine GPU (CPU-only)Apple M6 (unified CPU/GPU SoC, ARM64, AdvSIMD) · 24 GB RAM | lmstudio-mlxlmstudioMLX MXFP4 | Details → | |
| 11 | Llama-3.1-Nemotron-Nano-4B-v1.14.5BNVIDIA PerformancebenchmarkTimebench 3 - Kombi (Prefill + Generation) | 49,15 tok/s TG Prefill 1.978 · TTFT 1.328 ms | 1× | Keine GPU (CPU-only)Apple M6 (unified CPU/GPU SoC, ARM64, AdvSIMD) · 24 GB RAM | lmstudio-mlxlmstudioMLX 4bit | Details → | |
| 12 | Qwen3-4B4BQwen (Alibaba) PerformancebenchmarkTimebench 3 - Kombi (Prefill + Generation) | 48,34 tok/s TG Prefill 1.927 · TTFT 1.388 ms | 1× | Keine GPU (CPU-only)Apple M6 (unified CPU/GPU SoC, ARM64, AdvSIMD) · 24 GB RAM | lmstudio-mlxlmstudioMLX 4bit | Details → | |
| 13 | Gemma-4-26B-A4B-it26BGoogle PerformancebenchmarkTimebench 3 - Kombi (Prefill + Generation) | 47,35 tok/s TG Prefill 1.491 · TTFT 1.402 ms | 1× | Keine GPU (CPU-only)Apple M6 (unified CPU/GPU SoC, ARM64, AdvSIMD) · 24 GB RAM | lmstudio-mlxlmstudioMLX 4bit | Details → | |
| 14 | gemma-3n-E4B-it8BGoogle PerformancebenchmarkTimebench 3 - Kombi (Prefill + Generation) | 46,82 tok/s TG Prefill 1.560 · TTFT 1.426 ms | 1× | Keine GPU (CPU-only)Apple M6 (unified CPU/GPU SoC, ARM64, AdvSIMD) · 24 GB RAM | lmstudio-mlxlmstudioMLX 4bit | Details → | |
| 15 | Phi-3.5-mini-instruct3.8Bmicrosoft PerformancebenchmarkTimebench 3 - Kombi (Prefill + Generation) | 41,56 tok/s TG Prefill 1.815 · TTFT 1.555 ms | 1× | Keine GPU (CPU-only)Apple M6 (unified CPU/GPU SoC, ARM64, AdvSIMD) · 24 GB RAM | lmstudio-mlxlmstudioMLX 4bit | Details → | |
| 16 | SmolLM2-1.7B-Instruct1.7BHuggingFaceTB PerformancebenchmarkTimebench 3 - Kombi (Prefill + Generation) | 34,49 tok/s TG Prefill 4.031 · TTFT 886 ms | 1× | Keine GPU (CPU-only)Apple M6 (unified CPU/GPU SoC, ARM64, AdvSIMD) · 24 GB RAM | lmstudio-mlxlmstudioMLX bf16 | Details → | |
| 17 | Qwen2.5-7B-Instruct7BQwen (Alibaba) PerformancebenchmarkTimebench 3 - Kombi (Prefill + Generation) | 31,46 tok/s TG Prefill 1.123 · TTFT 2.322 ms | 1× | Keine GPU (CPU-only)Apple M6 (unified CPU/GPU SoC, ARM64, AdvSIMD) · 24 GB RAM | lmstudio-mlxlmstudioMLX 4bit | Details → | |
| 18 | Qwen2.5-Coder-7B-Instruct7.6BQwen (Alibaba) PerformancebenchmarkTimebench 3 - Kombi (Prefill + Generation) | 31,36 tok/s TG Prefill 1.117 · TTFT 2.334 ms | 1× | Keine GPU (CPU-only)Apple M6 (unified CPU/GPU SoC, ARM64, AdvSIMD) · 24 GB RAM | lmstudio-mlxlmstudioMLX 4bit | Details → | |
| 19 | Mamba-Codestral-7B-v0.17BMistral AI PerformancebenchmarkTimebench 3 - Kombi (Prefill + Generation) | 29,25 tok/s TG Prefill 327 · TTFT 8.424 ms | 1× | Keine GPU (CPU-only)Apple M6 (unified CPU/GPU SoC, ARM64, AdvSIMD) · 24 GB RAM | lmstudio-mlxlmstudioMLX 4bit | Details → | |
| 20 | Meta-Llama-3.1-8B-Instruct8BMeta PerformancebenchmarkTimebench 3 - Kombi (Prefill + Generation) | 29,20 tok/s TG Prefill 1.076 · TTFT 2.447 ms | 1× | Keine GPU (CPU-only)Apple M6 (unified CPU/GPU SoC, ARM64, AdvSIMD) · 24 GB RAM | lmstudio-mlxlmstudioMLX 4bit | Details → | |
| 21 | Qwen3-8B8.2BQwen (Alibaba) PerformancebenchmarkTimebench 3 - Kombi (Prefill + Generation) | 28,42 tok/s TG Prefill 1.096 · TTFT 2.410 ms | 1× | Keine GPU (CPU-only)Apple M6 (unified CPU/GPU SoC, ARM64, AdvSIMD) · 24 GB RAM | lmstudio-mlxlmstudioMLX 4bit | Details → | |
| 22 | gemma-2-9b-it9BGoogle PerformancebenchmarkTimebench 3 - Kombi (Prefill + Generation) | 21,98 tok/s TG Prefill 646 · TTFT 3.068 ms | 1× | Keine GPU (CPU-only)Apple M6 (unified CPU/GPU SoC, ARM64, AdvSIMD) · 24 GB RAM | lmstudio-mlxlmstudioMLX 4bit | Details → | |
| 23 | Ornith-1.5-9B9BDeepReinforce PerformancebenchmarkTimebench 3 - Kombi (Prefill + Generation) | 21,94 tok/s TG Prefill 1.047 · TTFT 2.009 ms | 1× | Keine GPU (CPU-only)Apple M6 (unified CPU/GPU SoC, ARM64, AdvSIMD) · 24 GB RAM | lmstudio-mlxlmstudioMLX 4bit (OptiQ) | Details → | |
| 24 | Ministral-3-14B-Reasoning-251214BMistral AI PerformancebenchmarkTimebench 3 - Kombi (Prefill + Generation) | 17,33 tok/s TG Prefill 621 · TTFT 3.879 ms | 1× | Keine GPU (CPU-only)Apple M6 (unified CPU/GPU SoC, ARM64, AdvSIMD) · 24 GB RAM | lmstudio-mlxlmstudioMLX 4bit | Details → | |
| 25 | Qwen3-14B14BQwen (Alibaba) PerformancebenchmarkTimebench 3 - Kombi (Prefill + Generation) | 15,53 tok/s TG Prefill 525 · TTFT 4.974 ms | 1× | Keine GPU (CPU-only)Apple M6 (unified CPU/GPU SoC, ARM64, AdvSIMD) · 24 GB RAM | lmstudio-mlxlmstudioMLX 4bit | Details → | |
| 26 | Mistral-Small-3.1-24B-Instruct-250324BMistral AI PerformancebenchmarkTimebench 3 - Kombi (Prefill + Generation) | 9,93 tok/s TG Prefill 290 · TTFT 8.335 ms | 1× | Keine GPU (CPU-only)Apple M6 (unified CPU/GPU SoC, ARM64, AdvSIMD) · 24 GB RAM | lmstudio-mlxlmstudioMLX 4bit | Details → | |
| 27 | Mistral-Small-24B-Instruct-250123.6BMistral AI PerformancebenchmarkTimebench 3 - Kombi (Prefill + Generation) | 9,90 tok/s TG Prefill 266 · TTFT 8.275 ms | 1× | Keine GPU (CPU-only)Apple M6 (unified CPU/GPU SoC, ARM64, AdvSIMD) · 24 GB RAM | lmstudio-mlxlmstudioMLX 4bit | Details → | |
| 28 | Devstral-Small-250724BMistral AI PerformancebenchmarkTimebench 3 - Kombi (Prefill + Generation) | 9,69 tok/s TG Prefill 427 · TTFT 8.728 ms | 1× | Keine GPU (CPU-only)Apple M6 (unified CPU/GPU SoC, ARM64, AdvSIMD) · 24 GB RAM | lmstudio-mlxlmstudioMLX 4bit | Details → | |
| 29 | Codestral-22B-v0.122BMistral AI PerformancebenchmarkTimebench 3 - Kombi (Prefill + Generation) | 9,65 tok/s TG Prefill 325 · TTFT 8.859 ms | 1× | Keine GPU (CPU-only)Apple M6 (unified CPU/GPU SoC, ARM64, AdvSIMD) · 24 GB RAM | lmstudio-mlxlmstudioMLX 4bit | Details → | |
| 30 | Devstral-Small-2-24B-Instruct-251224BMistral AI PerformancebenchmarkTimebench 3 - Kombi (Prefill + Generation) | 9,20 tok/s TG Prefill 261 · TTFT 8.641 ms | 1× | Keine GPU (CPU-only)Apple M6 (unified CPU/GPU SoC, ARM64, AdvSIMD) · 24 GB RAM | lmstudio-mlxlmstudioMLX 4bit | Details → |