🏆
Rankings
LLM Leaderboard
Real measurements of every tested language model on real, named hardware – rated by raw speed (performance in tokens per second, prefill and time to first token) and by practical task quality in complete agent and chat runs (harness). Pick a benchmark type below or filter by model, maker and hardware to see exactly what is tested and how the results are produced.
⚡ Performance (tok/s)🤖 Harness quality👥 Concurrency🖥️ real hardware
| # | Model / Maker | Metrics | Parallel | GPU / CPU / RAM | Runtime | ||
|---|---|---|---|---|---|---|---|
| 1 | gemma-4-E2B-it5BGoogle Performance benchmarkTimebench 3 - Kombi (Prefill + Generation) | 361,84 tok/s TG Prefill 7.233 · TTFT 2.750 ms | 10× | Tesla V100-PCIE-32GBAMD Ryzen 3 PRO 3200GE w/ Radeon Vega Graphics · 61 GB RAM | llama.cppgodclawQ8_0 | Details → | |
| 2 | gemma-4-E2B-it5BGoogle Performance benchmarkTimebench 3 - Kombi (Prefill + Generation) | 353,85 tok/s TG Prefill 4.758 · TTFT 4.353 ms | 10× | Tesla V100-PCIE-32GBAMD Ryzen 3 PRO 3200GE w/ Radeon Vega Graphics · 61 GB RAM | llama.cppgodclawQ4_K_M | Details → | |
| 3 | gemma-4-E2B-it5BGoogle Performance benchmarkTimebench 3 - Kombi (Prefill + Generation) | 275,62 tok/s TG Prefill 6.734 · TTFT 1.481 ms | 5× | Tesla V100-PCIE-32GBAMD Ryzen 3 PRO 3200GE w/ Radeon Vega Graphics · 61 GB RAM | llama.cppgodclawQ8_0 | Details → | |
| 4 | gemma-4-E2B-it5BGoogle Performance benchmarkTimebench 3 - Kombi (Prefill + Generation) | 255,41 tok/s TG Prefill 6.332 · TTFT 1.575 ms | 5× | Tesla V100-PCIE-32GBAMD Ryzen 3 PRO 3200GE w/ Radeon Vega Graphics · 61 GB RAM | llama.cppgodclawQ4_K_M | Details → | |
| 5 | Ornith-1.0-9B9BDeepReinforce Performance benchmarkTimebench 3 - Kombi (Prefill + Generation) | 248,72 tok/s TG Prefill 3.178 · TTFT 5.437 ms | 10× | Tesla V100-PCIE-32GBAMD Ryzen 3 PRO 3200GE w/ Radeon Vega Graphics · 61 GB RAM | llama.cppgodclawQ4_K_M | Details → | |
| 6 | Qwen3.6-35B-A3B35BQwen (Alibaba) Performance benchmarkTimebench 3 - Kombi (Prefill + Generation) | 225,93 tok/s TG Prefill 1.021 · TTFT 8.456 ms | 5× | Tesla V100-PCIE-32GBAMD Ryzen 3 PRO 3200GE w/ Radeon Vega Graphics · 61 GB RAM | llama.cppgodclawQ4_K_M | Details → | |
| 7 | Qwen3.6-35B-A3B35BQwen (Alibaba) Performance benchmarkTimebench 3 - Kombi (Prefill + Generation) | 216,62 tok/s TG Prefill 1.136 · TTFT 15.201 ms | 10× | Tesla V100-PCIE-32GBAMD Ryzen 3 PRO 3200GE w/ Radeon Vega Graphics · 61 GB RAM | llama.cppgodclawQ4_K_M | Details → | |
| 8 | gpt-oss-20b20BOpenAI Performance benchmarkTimebench 3 - Kombi (Prefill + Generation) | 206,60 tok/s TG Prefill 3.450 · TTFT 2.889 ms | 5× | Tesla V100-PCIE-32GBAMD Ryzen 3 PRO 3200GE w/ Radeon Vega Graphics · 61 GB RAM | llama.cppgodclawQ8_0 | Details → | |
| 9 | Ornith-1.0-9B9BDeepReinforce Performance benchmarkTimebench 3 - Kombi (Prefill + Generation) | 193,56 tok/s TG Prefill 3.011 · TTFT 2.862 ms | 5× | Tesla V100-PCIE-32GBAMD Ryzen 3 PRO 3200GE w/ Radeon Vega Graphics · 61 GB RAM | llama.cppgodclawQ4_K_M | Details → | |
| 10 | gemma-4-12B-it12BGoogle Performance benchmarkTimebench 3 - Kombi (Prefill + Generation) | 186,18 tok/s TG Prefill 1.748 · TTFT 10.113 ms | 10× | Tesla V100-PCIE-32GBAMD Ryzen 3 PRO 3200GE w/ Radeon Vega Graphics · 61 GB RAM | llama.cppgodclawQ4_K_M | Details → | |
| 11 | gemma-4-12B-it12BGoogle Performance benchmarkTimebench 3 - Kombi (Prefill + Generation) | 139,86 tok/s TG Prefill 1.753 · TTFT 5.039 ms | 5× | Tesla V100-PCIE-32GBAMD Ryzen 3 PRO 3200GE w/ Radeon Vega Graphics · 61 GB RAM | llama.cppgodclawQ4_K_M | Details → | |
| 12 | gpt-oss-20b20BOpenAI Performance benchmarkTimebench 3 - Kombi (Prefill + Generation) | 138,88 tok/s TG Prefill 2.354 · TTFT 9.328 ms | 10× | Tesla V100-PCIE-32GBAMD Ryzen 3 PRO 3200GE w/ Radeon Vega Graphics · 61 GB RAM | llama.cppgodclawQ8_0 | Details → | |
| 13 | gpt-oss-20b20BOpenAI Performance benchmarkTimebench 3 - Kombi (Prefill + Generation) | 132,28 tok/s TG Prefill 2.178 · TTFT 988 ms | 1× | Tesla V100-PCIE-32GBAMD Ryzen 3 PRO 3200GE w/ Radeon Vega Graphics · 61 GB RAM | llama.cppgodclawQ8_0 | Details → | |
| 14 | gemma-4-E2B-it5BGoogle Performance benchmarkTimebench 3 - Kombi (Prefill + Generation) | 128,96 tok/s TG Prefill 4.924 · TTFT 402 ms | 1× | Tesla V100-PCIE-32GBAMD Ryzen 3 PRO 3200GE w/ Radeon Vega Graphics · 61 GB RAM | llama.cppgodclawQ4_K_M | Details → | |
| 15 | Meta-Llama-3.1-8B-Instruct8BMeta Performance benchmarkTimebench 3 - Kombi (Prefill + Generation) | 122,55 tok/s TG Prefill 5.151 · TTFT 2.489 ms | 5× | Tesla V100-PCIE-32GBAMD Ryzen 3 PRO 3200GE w/ Radeon Vega Graphics · 61 GB RAM | llama.cppgodclawQ8_0 | Details → | |
| 16 | gemma-4-E2B-it5BGoogle Performance benchmarkTimebench 3 - Kombi (Prefill + Generation) | 117,98 tok/s TG Prefill 5.227 · TTFT 378 ms | 1× | Tesla V100-PCIE-32GBAMD Ryzen 3 PRO 3200GE w/ Radeon Vega Graphics · 61 GB RAM | llama.cppgodclawQ8_0 | Details → | |
| 17 | Qwen3.6-35B-A3B35BQwen (Alibaba) Performance benchmarkTimebench 3 - Kombi (Prefill + Generation) | 94,40 tok/s TG Prefill 768 · TTFT 2.243 ms | 1× | Tesla V100-PCIE-32GBAMD Ryzen 3 PRO 3200GE w/ Radeon Vega Graphics · 61 GB RAM | llama.cppgodclawQ4_K_M | Details → | |
| 18 | Ornith-1.0-9B9BDeepReinforce Performance benchmarkTimebench 3 - Kombi (Prefill + Generation) | 93,40 tok/s TG Prefill 3.060 · TTFT 3.167 ms | 5× | Tesla V100-PCIE-32GBAMD Ryzen 3 PRO 3200GE w/ Radeon Vega Graphics · 61 GB RAM | llama.cppgodclawQ8_0 | Details → | |
| 19 | Ornith-1.0-9B9BDeepReinforce Performance benchmarkTimebench 3 - Kombi (Prefill + Generation) | 83,75 tok/s TG Prefill 1.512 · TTFT 1.139 ms | 1× | Tesla V100-PCIE-32GBAMD Ryzen 3 PRO 3200GE w/ Radeon Vega Graphics · 61 GB RAM | llama.cppgodclawQ4_K_M | Details → | |
| 20 | Meta-Llama-3.1-8B-Instruct8BMeta Performance benchmarkTimebench 3 - Kombi (Prefill + Generation) | 83,10 tok/s TG Prefill 6.957 · TTFT 7.450 ms | 10× | Tesla V100-PCIE-32GBAMD Ryzen 3 PRO 3200GE w/ Radeon Vega Graphics · 61 GB RAM | llama.cppgodclawQ8_0 | Details → | |
| 21 | Ornith-1.0-9B9BDeepReinforce Performance benchmarkTimebench 3 - Kombi (Prefill + Generation) | 74,39 tok/s TG Prefill 1.748 · TTFT 11.355 ms | 10× | Tesla V100-PCIE-32GBAMD Ryzen 3 PRO 3200GE w/ Radeon Vega Graphics · 61 GB RAM | llama.cppgodclawQ8_0 | Details → | |
| 22 | Meta-Llama-3.1-8B-Instruct8BMeta Performance benchmarkTimebench 3 - Kombi (Prefill + Generation) | 69,93 tok/s TG Prefill 2.843 · TTFT 920 ms | 1× | Tesla V100-PCIE-32GBAMD Ryzen 3 PRO 3200GE w/ Radeon Vega Graphics · 61 GB RAM | llama.cppgodclawQ8_0 | Details → | |
| 23 | Ornith-1.0-9B9BDeepReinforce Performance benchmarkTimebench 3 - Kombi (Prefill + Generation) | 62,07 tok/s TG Prefill 2.204 · TTFT 932 ms | 1× | Tesla V100-PCIE-32GBAMD Ryzen 3 PRO 3200GE w/ Radeon Vega Graphics · 61 GB RAM | llama.cppgodclawQ8_0 | Details → | |
| 24 | gemma-4-12B-it12BGoogle Performance benchmarkTimebench 3 - Kombi (Prefill + Generation) | 58,59 tok/s TG Prefill 1.147 · TTFT 1.541 ms | 1× | Tesla V100-PCIE-32GBAMD Ryzen 3 PRO 3200GE w/ Radeon Vega Graphics · 61 GB RAM | llama.cppgodclawQ4_K_M | Details → | |
| 25 | Ministral-3-14B-Reasoning-251214BMistral AI Performance benchmarkTimebench 3 - Kombi (Prefill + Generation) | 52,78 tok/s TG Prefill 1.962 · TTFT 6.490 ms | 5× | Tesla V100-PCIE-32GBAMD Ryzen 3 PRO 3200GE w/ Radeon Vega Graphics · 61 GB RAM | llama.cppgodclawQ8_0 | Details → | |
| 26 | Ministral-3-14B-Reasoning-251214BMistral AI Performance benchmarkTimebench 3 - Kombi (Prefill + Generation) | 52,41 tok/s TG Prefill 1.260 · TTFT 19.903 ms | 10× | Tesla V100-PCIE-32GBAMD Ryzen 3 PRO 3200GE w/ Radeon Vega Graphics · 61 GB RAM | llama.cppgodclawQ8_0 | Details → | |
| 27 | gemma-4-12B-it12BGoogle Performance benchmarkTimebench 3 - Kombi (Prefill + Generation) | 49,35 tok/s TG Prefill 778 · TTFT 25.706 ms | 10× | Tesla V100-PCIE-32GBAMD Ryzen 3 PRO 3200GE w/ Radeon Vega Graphics · 61 GB RAM | llama.cppgodclawQ8_0 | Details → | |
| 28 | gemma-4-12B-it12BGoogle Performance benchmarkTimebench 3 - Kombi (Prefill + Generation) | 47,87 tok/s TG Prefill 1.491 · TTFT 7.113 ms | 5× | Tesla V100-PCIE-32GBAMD Ryzen 3 PRO 3200GE w/ Radeon Vega Graphics · 61 GB RAM | llama.cppgodclawQ8_0 | Details → | |
| 29 | gemma-4-12B-it12BGoogle Performance benchmarkTimebench 3 - Kombi (Prefill + Generation) | 36,37 tok/s TG Prefill 928 · TTFT 13.981 ms | 1× | Tesla V100-PCIE-32GBAMD Ryzen 3 PRO 3200GE w/ Radeon Vega Graphics · 61 GB RAM | llama.cppgodclawQ8_0 | Details → | |
| 30 | Ministral-3-14B-Reasoning-251214BMistral AI Performance benchmarkTimebench 3 - Kombi (Prefill + Generation) | 35,13 tok/s TG Prefill 2.320 · TTFT 975 ms | 1× | Tesla V100-PCIE-32GBAMD Ryzen 3 PRO 3200GE w/ Radeon Vega Graphics · 61 GB RAM | llama.cppgodclawQ8_0 | Details → |