🏆 Rangliste
Highscore
Die besten Laeufe im direkten Ranking - umschaltbar zwischen Performance (Token/s) und Harness (Aufgabenpunkte).
| # | Modell / Hersteller | Grafikkarte | CPU / RAM | Erfolgsquote | Punkte | Aufgaben | Parallel | Runtime | |
|---|---|---|---|---|---|---|---|---|---|
| 🥇 | Gemma-4-26B-A4B-itGoogle | NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition | 32x AMD Ryzen 9 9950X 16-Core Processor 92 GB | 91% |
2.701 / 2.980 | 16 / 16 | 1× | openclaw_cli | Details → |
| 🥈 | Devstral-Small-2-24B-Instruct-2512Mistral AI | AMD GPU | 32x AMD RYZEN AI MAX+ 395 w/ Radeon 8060S 31 GB | 87% |
2.597 / 2.980 | 16 / 16 | 1× | llama.cppgodclawQ8_0 | Details → |
| 🥉 | Gemma-4-26B-A4B-itGoogle | AMD GPU | 32x AMD RYZEN AI MAX+ 395 w/ Radeon 8060S 123 GB | 82% |
2.440 / 2.980 | 16 / 16 | 1× | openclaw_cli | Details → |
| 4 | MiniMax-M2.5MiniMax | 3x NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition | 34x AMD Ryzen Threadripper PRO 9965WX 24-Cores 125 GB | 81% |
2.427 / 2.980 | 16 / 16 | 1× | vLLMgodclawBF16 | Details → |
| 5 | Gemma-4-26B-A4B-itGoogle | AMD GPU | 32x AMD RYZEN AI MAX+ 395 w/ Radeon 8060S 123 GB | 80% |
2.383 / 2.980 | 16 / 16 | 1× | openclaw_cli | Details → |
| 6 | MiniMax-M2.5MiniMax | 3x NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition | 34x AMD Ryzen Threadripper PRO 9965WX 24-Cores 125 GB | 79% |
2.340 / 2.980 | 16 / 16 | 1× | openclaw_cli | Details → |
| 7 | Gemma-4-26B-A4B-itGoogle | AMD GPU | 32x AMD RYZEN AI MAX+ 395 w/ Radeon 8060S 31 GB | 76% |
2.268 / 2.980 | 16 / 16 | 1× | llama.cppgodclawQ8_0 | Details → |
| 8 | Devstral-Small-2-24B-Instruct-2512Mistral AI | AMD GPU | 32x AMD RYZEN AI MAX+ 395 w/ Radeon 8060S 31 GB | 76% |
5.738 / 7.580 | 40 / 40 | 1× | llama.cppgodclawQ8_0 | Details → |
| 9 | Gemma-4-26B-A4B-itGoogle | AMD GPU | 32x AMD RYZEN AI MAX+ 395 w/ Radeon 8060S 31 GB | 74% |
2.197 / 2.980 | 16 / 16 | 1× | openclaw_cli | Details → |
| 10 | MiniMax-M2.5-BF16-INT4-AWQMiniMax | NVIDIA GB10 (DGX Spark) | NVIDIA Grace 120 GB | 71% |
1.245 / 1.760 | 15 / 15 | 1× | openclaw_cliBF16 | Details → |
| 11 | Devstral-Small-2-24B-Instruct-2512Mistral AI | AMD GPU | 32x AMD RYZEN AI MAX+ 395 w/ Radeon 8060S 31 GB | 69% |
5.629 / 8.210 | 40 / 40 | 1× | llama.cppgodclawQ8_0 | Details → |
| 12 | MiniMax-M2.5MiniMax | 3x NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition | 34x AMD Ryzen Threadripper PRO 9965WX 24-Cores 125 GB | 68% |
2.033 / 2.980 | 16 / 16 | 1× | vLLMgodclawBF16 | Details → |
| 13 | MiniMax-M2.5MiniMax | 3x NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition | 34x AMD Ryzen Threadripper PRO 9965WX 24-Cores 125 GB | 65% |
130 / 200 | 2 / 2 | 1× | godclaw | Details → |
| 14 | MiniMax-M2.5MiniMax | 3x NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition | 34x AMD Ryzen Threadripper PRO 9965WX 24-Cores 125 GB | 61% |
1.079 / 1.760 | 15 / 15 | 1× | openclaw_cli | Details → |
| 15 | MiniMax-M2.7-AWQ-4bitMiniMax | NVIDIA GB10 (DGX Spark) | NVIDIA Grace 120 GB | 61% |
3.641 / 5.960 | 40 / 40 | 1× | vLLMgodclawAWQ | Details → |
| 16 | MiniMax-M2.5MiniMax | 3x NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition | 34x AMD Ryzen Threadripper PRO 9965WX 24-Cores 125 GB | 61% |
3.622 / 5.960 | 34 / 40 | 1× | vLLMgodclawBF16 | Details → |
| 17 | MiniMax-M2.5-BF16-INT4-AWQMiniMax | NVIDIA GB10 (DGX Spark) | NVIDIA Grace 120 GB | 59% |
1.762 / 2.980 | 16 / 16 | 1× | openclaw_cliBF16 | Details → |
| 18 | MiniMax-M2.5MiniMax | 3x NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition | 34x AMD Ryzen Threadripper PRO 9965WX 24-Cores 125 GB | 58% |
1.714 / 2.980 | 16 / 16 | 1× | vLLMgodclawBF16 | Details → |
| 19 | MiniMax-M2.5MiniMax | 3x NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition | 34x AMD Ryzen Threadripper PRO 9965WX 24-Cores 125 GB | 53% |
106 / 200 | 2 / 2 | 1× | openclaw_cli | Details → |
| 20 | MiniMax-M2.5-BF16-INT4-AWQMiniMax | NVIDIA GB10 (DGX Spark) | NVIDIA Grace 120 GB | 52% |
790 / 1.520 | 15 / 15 | 1× | openclaw_cliBF16 | Details → |
| 21 | Qwen3.6-27BQwen (Alibaba) | NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition | 32x AMD Ryzen 9 9950X 16-Core Processor 92 GB | 41% |
1.220 / 2.980 | 16 / 16 | 1× | vLLMgodclaw | Details → |
| 22 | Qwen3.6-27BQwen (Alibaba) | NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition | 32x AMD Ryzen 9 9950X 16-Core Processor 92 GB | 38% |
2.267 / 5.960 | 38 / 40 | 1× | vLLMgodclaw | Details → |
| 23 | Gemma-4-26B-A4B-itGoogle | AMD GPU | 32x AMD RYZEN AI MAX+ 395 w/ Radeon 8060S 31 GB | 30% |
2.271 / 7.580 | 18 / 40 | 1× | llama.cppgodclawQ8_0 | Details → |
| 24 | MiniMax-M2.7-AWQ-4bitMiniMax | NVIDIA GB10 (DGX Spark) | NVIDIA Grace 120 GB | 29% |
858 / 2.980 | 16 / 16 | 1× | godclawAWQ | Details → |
| 25 | Gemma-4-26B-A4B-itGoogle | AMD GPU | 32x AMD RYZEN AI MAX+ 395 w/ Radeon 8060S 31 GB | 25% |
1.489 / 5.960 | 22 / 40 | 1× | llama.cppgodclawQ8_0 | Details → |
| 26 | Devstral-Small-2-24B-Instruct-2512Mistral AI | AMD GPU | 32x AMD RYZEN AI MAX+ 395 w/ Radeon 8060S 31 GB | 24% |
1.436 / 5.960 | 21 / 40 | 1× | llama.cppgodclawQ8_0 | Details → |
| 27 | Qwen2.5-32B-Instruct-AWQQwen (Alibaba) | AMD Radeon 8060S Graphics | 32x AMD RYZEN AI MAX+ 395 w/ Radeon 8060S 31 GB | 13% |
235 / 1.760 | 15 / 15 | 1× | openclaw_cliAWQ | Details → |
| 28 | Qwen2.5-32B-Instruct-AWQQwen (Alibaba) | AMD GPU | 32x AMD RYZEN AI MAX+ 395 w/ Radeon 8060S 123 GB | 10% |
298 / 2.980 | 16 / 16 | 1× | openclaw_cliAWQ | Details → |
| 29 | gemma4:31bGoogle | AMD GPU | 32x AMD RYZEN AI MAX+ 395 w/ Radeon 8060S 123 GB | 10% |
298 / 2.980 | 16 / 16 | 1× | openclaw_cli | Details → |
| 30 | MiniMax-M2.5MiniMax | 3x NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition | 34x AMD Ryzen Threadripper PRO 9965WX 24-Cores 125 GB | 10% |
10 / 100 | 1 / 1 | 1× | openclaw_cli | Details → |
Ranking aus veroeffentlichten Benchmarks. Sortiert nach Erfolgsquote, dann erreichten Punkten. Wie wird gemessen?