🏆 Rangliste
Highscore
Die besten Laeufe im direkten Ranking - umschaltbar zwischen Performance (Token/s) und Harness (Aufgabenpunkte).
| # | Modell / Hersteller | Grafikkarte | CPU / RAM | Prefill | Prefill | TG | TTFT | Parallel | Runtime | |
|---|---|---|---|---|---|---|---|---|---|---|
| 🥇 | Qwen3-30B-A3B-Instruct-2507Qwen (Alibaba) | NVIDIA GeForce RTX 5090 | 16x AMD Ryzen 7 5800X3D 8-Core Processor 126 GB | 37.116,1 tok/s |
37.116 | 909,3 | 1783 ms | 10× | vLLMcodex_cliAWQ | Details → |
| 🥈 | gpt-oss-20bOpenAI | NVIDIA RTX PRO 6000 Blackwell Workstation Edition | 32x AMD Ryzen 9 9950X 16-Core Processor 92 GB | 35.347,7 tok/s |
35.348 | 1.443,9 | 560 ms | 10× | vLLMgodclaw | Details → |
| 🥉 | Qwen3-Coder-30B-A3B-InstructQwen (Alibaba) | NVIDIA GeForce RTX 5090 | 16x AMD Ryzen 7 5800X3D 8-Core Processor 126 GB | 33.767,8 tok/s |
33.768 | 966,2 | 1792 ms | 10× | vLLMcodex_cliAWQ | Details → |
| 4 | Qwen3-30B-A3B-Thinking-2507Qwen (Alibaba) | NVIDIA GeForce RTX 5090 | 16x AMD Ryzen 7 5800X3D 8-Core Processor 126 GB | 33.022,6 tok/s |
33.023 | 873,7 | 1834 ms | 10× | vLLMcodex_cliAWQ | Details → |
| 5 | gpt-oss-20bOpenAI | NVIDIA RTX PRO 6000 Blackwell Workstation Edition | 32x AMD Ryzen 9 9950X 16-Core Processor 92 GB | 31.097,5 tok/s |
31.098 | 842,9 | 318 ms | 5× | vLLMgodclaw | Details → |
| 6 | gpt-oss-20bOpenAI | NVIDIA GeForce RTX 5090 | 16x AMD Ryzen 7 5800X3D 8-Core Processor 126 GB | 25.461,8 tok/s |
25.462 | 1.295,3 | 808 ms | 10× | vLLMcodex_cliAWQ | Details → |
| 7 | NVIDIA-Nemotron-3-Nano-30B-A3B-BF16NVIDIA | 3x AMD Radeon AI PRO R9700 | 32x AMD Ryzen Threadripper PRO 7955WX 16-Cores 184 GB | 23.508,3 tok/s |
23.508 | 153,3 | 1416 ms | 10× | vLLMgodclawBF16 | Details → |
| 8 | Qwen3-30B-A3B-Thinking-2507Qwen (Alibaba) | NVIDIA GeForce RTX 5090 | 16x AMD Ryzen 7 5800X3D 8-Core Processor 126 GB | 20.291,9 tok/s |
20.292 | 633,4 | 767 ms | 5× | vLLMcodex_cliAWQ | Details → |
| 9 | Ministral-3-14B-Reasoning-2512Mistral AI | NVIDIA RTX PRO 6000 Blackwell Workstation Edition | 32x AMD Ryzen 9 9950X 16-Core Processor 92 GB | 20.061,1 tok/s |
20.061 | 1.015,7 | 1288 ms | 10× | vLLMgodclawAWQ | Details → |
| 10 | Qwen3-30B-A3B-Instruct-2507Qwen (Alibaba) | NVIDIA GeForce RTX 5090 | 16x AMD Ryzen 7 5800X3D 8-Core Processor 126 GB | 17.990,2 tok/s |
17.990 | 655,8 | 930 ms | 5× | vLLMcodex_cliAWQ | Details → |
| 11 | gpt-oss-20bOpenAI | NVIDIA GeForce RTX 5090 | 16x AMD Ryzen 7 5800X3D 8-Core Processor 126 GB | 17.724,5 tok/s |
17.724 | 784,5 | 560 ms | 5× | vLLMcodex_cliAWQ | Details → |
| 12 | gpt-oss-20bOpenAI | NVIDIA RTX PRO 6000 Blackwell Workstation Edition | 32x AMD Ryzen 9 9950X 16-Core Processor 92 GB | 17.370,9 tok/s |
17.371 | 268,7 | 113 ms | 1× | vLLMgodclaw | Details → |
| 13 | Qwen3-Coder-30B-A3B-InstructQwen (Alibaba) | NVIDIA GeForce RTX 5090 | 16x AMD Ryzen 7 5800X3D 8-Core Processor 126 GB | 16.191,7 tok/s |
16.192 | 688,6 | 873 ms | 5× | vLLMcodex_cliAWQ | Details → |
| 14 | Ministral-3-14B-Reasoning-2512Mistral AI | NVIDIA RTX PRO 6000 Blackwell Workstation Edition | 32x AMD Ryzen 9 9950X 16-Core Processor 92 GB | 13.999,4 tok/s |
13.999 | 615,8 | 796 ms | 5× | vLLMgodclawAWQ | Details → |
| 15 | Qwen3-Omni-30B-A3B-ThinkingQwen (Alibaba) | 3x AMD Radeon AI PRO R9700 | 32x AMD Ryzen Threadripper PRO 7955WX 16-Cores 184 GB | 13.527,8 tok/s |
13.528 | 160,6 | 3189 ms | 10× | vLLMgodclawAWQ | Details → |
| 16 | gpt-oss-20bOpenAI | NVIDIA GeForce RTX 3090 Ti | 32x AMD Ryzen 9 8945HX with Radeon Graphics 92 GB | 11.891,9 tok/s |
11.892 | 796,8 | 1780 ms | 10× | vLLMwebsocket | Details → |
| 17 | Qwen3-Coder-30B-A3B-InstructQwen (Alibaba) | 3x AMD Radeon AI PRO R9700 | 32x AMD Ryzen Threadripper PRO 7955WX 16-Cores 184 GB | 11.379,2 tok/s |
11.379 | 237,2 | 2588 ms | 10× | llama.cppgodclawQ4_K_M | Details → |
| 18 | Qwen3-Omni-30B-A3B-ThinkingQwen (Alibaba) | 3x AMD Radeon AI PRO R9700 | 32x AMD Ryzen Threadripper PRO 7955WX 16-Cores 184 GB | 10.572,4 tok/s |
10.572 | 100,3 | 1487 ms | 5× | vLLMgodclawAWQ | Details → |
| 19 | NVIDIA-Nemotron-3-Nano-30B-A3B-BF16NVIDIA | 3x AMD Radeon AI PRO R9700 | 32x AMD Ryzen Threadripper PRO 7955WX 16-Cores 184 GB | 10.458,4 tok/s |
10.458 | 102,5 | 1707 ms | 5× | vLLMgodclawBF16 | Details → |
| 20 | Laguna-XS-2.1-INT4Poolside | 3x AMD Radeon AI PRO R9700 | 32x AMD Ryzen Threadripper PRO 7955WX 16-Cores 184 GB | 10.264,4 tok/s |
10.264 | 154,3 | 3945 ms | 10× | vLLMgodclawINT4 | Details → |
| 21 | Ministral-3-14B-Reasoning-2512Mistral AI | NVIDIA RTX PRO 6000 Blackwell Workstation Edition | 32x AMD Ryzen 9 9950X 16-Core Processor 92 GB | 9.895,0 tok/s |
9.895 | 144,8 | 216 ms | 1× | vLLMgodclawAWQ | Details → |
| 22 | gpt-oss-20bOpenAI | NVIDIA GeForce RTX 3090 Ti | 32x AMD Ryzen 9 8945HX with Radeon Graphics 92 GB | 9.270,2 tok/s |
9.270 | 483,7 | 1069 ms | 5× | vLLMwebsocket | Details → |
| 23 | Mistral-Small-3.1-24B-Instruct-2503Mistral AI | 3x AMD Radeon AI PRO R9700 | 32x AMD Ryzen Threadripper PRO 7955WX 16-Cores 184 GB | 9.072,4 tok/s |
9.072 | 15,0 | 244 ms | 1× | vLLMgodclawAWQ | Details → |
| 24 | Laguna-S-2.1-INT4Poolside | 3x AMD Radeon AI PRO R9700 | 32x AMD Ryzen Threadripper PRO 7955WX 16-Cores 184 GB | 8.629,4 tok/s |
8.629 | 13,4 | 304 ms | 1× | vLLMgodclawINT4 | Details → |
| 25 | Qwen3-Coder-30B-A3B-InstructQwen (Alibaba) | 3x AMD Radeon AI PRO R9700 | 32x AMD Ryzen Threadripper PRO 7955WX 16-Cores 184 GB | 8.401,2 tok/s |
8.401 | 205,0 | 1506 ms | 5× | llama.cppgodclawQ4_K_M | Details → |
| 26 | Qwen3-30B-A3B-Thinking-2507Qwen (Alibaba) | NVIDIA GeForce RTX 5090 | 16x AMD Ryzen 7 5800X3D 8-Core Processor 126 GB | 8.379,2 tok/s |
8.379 | 221,3 | 300 ms | 1× | vLLMcodex_cliAWQ | Details → |
| 27 | Qwen3-Coder-30B-A3B-InstructQwen (Alibaba) | 3x AMD Radeon AI PRO R9700 | 32x AMD Ryzen Threadripper PRO 7955WX 16-Cores 184 GB | 8.376,7 tok/s |
8.377 | 38,3 | 386 ms | 1× | vLLMgodclawAWQ | Details → |
| 28 | Qwen3-30B-A3B-Instruct-2507Qwen (Alibaba) | NVIDIA GeForce RTX 5090 | 16x AMD Ryzen 7 5800X3D 8-Core Processor 126 GB | 8.323,3 tok/s |
8.323 | 232,8 | 311 ms | 1× | vLLMcodex_cliAWQ | Details → |
| 29 | Qwen2.5-Coder-32B-InstructQwen (Alibaba) | 3x AMD Radeon AI PRO R9700 | 32x AMD Ryzen Threadripper PRO 7955WX 16-Cores 184 GB | 8.243,6 tok/s |
8.244 | 9,9 | 264 ms | 1× | vLLMgodclawAWQ | Details → |
| 30 | gpt-oss-20bOpenAI | NVIDIA GeForce RTX 3090 Ti | 32x AMD Ryzen 9 8945HX with Radeon Graphics 92 GB | 8.073,6 tok/s |
8.074 | 176,2 | 245 ms | 1× | vLLMwebsocket | Details → |
| 31 | Qwen3-30B-A3B-Thinking-2507Qwen (Alibaba) | 3x AMD Radeon AI PRO R9700 | 32x AMD Ryzen Threadripper PRO 7955WX 16-Cores 184 GB | 8.044,0 tok/s |
8.044 | 42,0 | 377 ms | 1× | vLLMgodclawAWQ | Details → |
| 32 | Qwen3-VL-30B-A3B-InstructQwen (Alibaba) | 3x AMD Radeon AI PRO R9700 | 32x AMD Ryzen Threadripper PRO 7955WX 16-Cores 184 GB | 7.938,6 tok/s |
7.939 | 36,9 | 472 ms | 1× | vLLMgodclawAWQ | Details → |
| 33 | Qwen3-VL-30B-A3B-InstructQwen (Alibaba) | 3x AMD Radeon AI PRO R9700 | 32x AMD Ryzen Threadripper PRO 7955WX 16-Cores 184 GB | 7.861,3 tok/s |
7.861 | 34,4 | 473 ms | 1× | vLLMgodclawAWQ | Details → |
| 34 | Ministral-3-14B-Reasoning-2512Mistral AI | NVIDIA GeForce RTX 3090 Ti | 32x AMD Ryzen 9 8945HX with Radeon Graphics 92 GB | 7.751,4 tok/s |
7.751 | 524,4 | 4130 ms | 10× | vLLMwebsocketAWQ | Details → |
| 35 | Laguna-XS-2.1-INT4Poolside | 3x AMD Radeon AI PRO R9700 | 32x AMD Ryzen Threadripper PRO 7955WX 16-Cores 184 GB | 7.618,0 tok/s |
7.618 | 88,9 | 2213 ms | 5× | vLLMgodclawINT4 | Details → |
| 36 | Qwen2.5-Coder-32B-InstructQwen (Alibaba) | 3x AMD Radeon AI PRO R9700 | 32x AMD Ryzen Threadripper PRO 7955WX 16-Cores 184 GB | 7.325,1 tok/s |
7.325 | 9,8 | 310 ms | 1× | vLLMgodclawAWQ | Details → |
| 37 | Qwen3-VL-30B-A3B-InstructQwen (Alibaba) | 3x AMD Radeon AI PRO R9700 | 32x AMD Ryzen Threadripper PRO 7955WX 16-Cores 184 GB | 7.288,9 tok/s |
7.289 | 69,2 | 412 ms | 1× | vLLMgodclawAWQ | Details → |
| 38 | gpt-oss-20bOpenAI | NVIDIA GeForce RTX 5090 | 16x AMD Ryzen 7 5800X3D 8-Core Processor 126 GB | 7.272,7 tok/s |
7.273 | 267,0 | 271 ms | 1× | vLLMcodex_cliAWQ | Details → |
| 39 | North-Mini-Code-1.0Cohere | 3x AMD Radeon AI PRO R9700 | 32x AMD Ryzen Threadripper PRO 7955WX 16-Cores 184 GB | 6.997,2 tok/s |
6.997 | 279,0 | 3294 ms | 10× | llama.cppgodclawQ4_K_M | Details → |
| 40 | Qwen3-30B-A3B-Instruct-2507Qwen (Alibaba) | 3x AMD Radeon AI PRO R9700 | 32x AMD Ryzen Threadripper PRO 7955WX 16-Cores 184 GB | 6.737,2 tok/s |
6.737 | 173,8 | 4318 ms | 10× | llama.cppgodclawQ4_K_M | Details → |
| 41 | gpt-oss-20bOpenAI | 3x AMD Radeon AI PRO R9700 | 32x AMD Ryzen Threadripper PRO 7955WX 16-Cores 184 GB | 6.539,7 tok/s |
6.540 | 35,7 | 374 ms | 1× | vLLMgodclaw4-Bit | Details → |
| 42 | Qwen3-VL-30B-A3B-InstructQwen (Alibaba) | 3x AMD Radeon AI PRO R9700 | 32x AMD Ryzen Threadripper PRO 7955WX 16-Cores 184 GB | 6.280,9 tok/s |
6.281 | 167,2 | 4625 ms | 10× | llama.cppgodclawQ4_K_M | Details → |
| 43 | Qwen3.6-35B-A3BQwen (Alibaba) | 3x AMD Radeon AI PRO R9700 | 32x AMD Ryzen Threadripper PRO 7955WX 16-Cores 184 GB | 5.971,1 tok/s |
5.971 | 34,2 | 356 ms | 1× | vLLMgodclawAWQ | Details → |
| 44 | Laguna-S-2.1-INT4Poolside | 3x AMD Radeon AI PRO R9700 | 32x AMD Ryzen Threadripper PRO 7955WX 16-Cores 184 GB | 5.970,3 tok/s |
5.970 | 70,5 | 7474 ms | 10× | vLLMgodclawINT4 | Details → |
| 45 | Ministral-3-14B-Reasoning-2512Mistral AI | NVIDIA GeForce RTX 3090 Ti | 32x AMD Ryzen 9 8945HX with Radeon Graphics 92 GB | 5.943,3 tok/s |
5.943 | 338,3 | 2299 ms | 5× | vLLMwebsocketAWQ | Details → |
| 46 | Qwen3-Coder-30B-A3B-InstructQwen (Alibaba) | 3x AMD Radeon AI PRO R9700 | 32x AMD Ryzen Threadripper PRO 7955WX 16-Cores 184 GB | 5.913,2 tok/s |
5.913 | 47,7 | 437 ms | 1× | vLLMgodclawAWQ | Details → |
| 47 | NVIDIA-Nemotron-3-Nano-30B-A3B-BF16NVIDIA | 3x AMD Radeon AI PRO R9700 | 32x AMD Ryzen Threadripper PRO 7955WX 16-Cores 184 GB | 5.838,9 tok/s |
5.839 | 38,3 | 376 ms | 1× | vLLMgodclawBF16 | Details → |
| 48 | Nemotron-Cascade-2-30B-A3BNVIDIA | 3x AMD Radeon AI PRO R9700 | 32x AMD Ryzen Threadripper PRO 7955WX 16-Cores 184 GB | 5.684,3 tok/s |
5.684 | 270,4 | 3915 ms | 10× | llama.cppgodclawQ4_K_M | Details → |
| 49 | Qwen3-Coder-30B-A3B-InstructQwen (Alibaba) | 3x AMD Radeon AI PRO R9700 | 32x AMD Ryzen Threadripper PRO 7955WX 16-Cores 184 GB | 5.663,6 tok/s |
5.664 | 40,5 | 755 ms | 1× | vLLMgodclawAWQ | Details → |
| 50 | MiniMax-M2.7-AWQ-4bitMiniMax | NVIDIA GB10 (DGX Spark) | NVIDIA Grace 120 GB | 5.634,4 tok/s |
5.634 | 106,1 | 12854 ms | 5× | vLLMgodclawAWQ | Details → |
Ranking aus veroeffentlichten Benchmarks. Sortiert nach Prefill-Durchsatz. Wie wird gemessen?