🏆
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 | Qwen3-0.6B0.6BQwen (Alibaba) Performance benchmarkTimebench 3 - Kombi (Prefill + Generation) | 503,33 tok/s TG Prefill 112.346 · TTFT 242 ms | 10× | NVIDIA GeForce RTX 4090AMD Ryzen 3 3100 4-Core Processor · 61 GB RAM | vLLMopenclaw_cliFP8 | Details → | |
| 2 | Meta-Llama-3.1-8B-Instruct8BMeta Performance benchmarkTimebench 3 - Kombi (Prefill + Generation) | 489,72 tok/s TG Prefill 21.352 · TTFT 2.183 ms | 10× | NVIDIA GeForce RTX 4090AMD Ryzen 3 3100 4-Core Processor · 61 GB RAM | vLLMopenclaw_cliAWQ | Details → | |
| 3 | Qwen3-Coder-30B-A3B-Instruct30BQwen (Alibaba) Performance benchmarkTimebench 3 - Kombi (Prefill + Generation) | 400,42 tok/s TG Prefill 12.694 · TTFT 1.086 ms | 5× | NVIDIA GeForce RTX 4090AMD Ryzen 3 3100 4-Core Processor · 61 GB RAM | llama.cppopenclaw_cliQ4_K_M | Details → | |
| 4 | Devstral-Small-2-24B-Instruct-251224BMistral AI Performance benchmarkTimebench 3 - Kombi (Prefill + Generation) | 265,94 tok/s TG Prefill 8.432 · TTFT 3.619 ms | 10× | NVIDIA GeForce RTX 4090AMD Ryzen 3 3100 4-Core Processor · 61 GB RAM | llama.cppopenclaw_cliQ4_K_M | Details → | |
| 5 | Qwen3-0.6B0.6BQwen (Alibaba) Performance benchmarkTimebench 3 - Kombi (Prefill + Generation) | 258,23 tok/s TG Prefill 81.661 · TTFT 159 ms | 5× | NVIDIA GeForce RTX 4090AMD Ryzen 3 3100 4-Core Processor · 61 GB RAM | vLLMopenclaw_cliFP8 | Details → | |
| 6 | Meta-Llama-3.1-8B-Instruct8BMeta Performance benchmarkTimebench 3 - Kombi (Prefill + Generation) | 254,69 tok/s TG Prefill 14.644 · TTFT 1.771 ms | 5× | NVIDIA GeForce RTX 4090AMD Ryzen 3 3100 4-Core Processor · 61 GB RAM | vLLMopenclaw_cliAWQ | Details → | |
| 7 | Qwen3-Coder-30B-A3B-Instruct30BQwen (Alibaba) Performance benchmarkTimebench 3 - Kombi (Prefill + Generation) | 198,88 tok/s TG Prefill 7.188 · TTFT 339 ms | 1× | NVIDIA GeForce RTX 4090AMD Ryzen 3 3100 4-Core Processor · 61 GB RAM | llama.cppopenclaw_cliQ4_K_M | Details → | |
| 8 | Devstral-Small-2-24B-Instruct-251224BMistral AI Performance benchmarkTimebench 3 - Kombi (Prefill + Generation) | 186,91 tok/s TG Prefill 6.423 · TTFT 2.079 ms | 5× | NVIDIA GeForce RTX 4090AMD Ryzen 3 3100 4-Core Processor · 61 GB RAM | llama.cppopenclaw_cliQ4_K_M | Details → | |
| 9 | Ministral-3-14B-Reasoning-251214BMistral AI Performance benchmarkTimebench 3 - Kombi (Prefill + Generation) | 157,87 tok/s TG Prefill 8.091 · TTFT 1.495 ms | 5× | NVIDIA GeForce RTX 4090AMD Ryzen 3 3100 4-Core Processor · 61 GB RAM | llama.cppopenclaw_cliQ8_0 | Details → | |
| 10 | Qwen3-Coder-30B-A3B-Instruct30BQwen (Alibaba) Performance benchmarkTimebench 3 - Kombi (Prefill + Generation) | 68,82 tok/s TG Prefill 3.989 · TTFT 7.885 ms | 10× | NVIDIA GeForce RTX 4090AMD Ryzen 3 3100 4-Core Processor · 61 GB RAM | llama.cppopenclaw_cliQ4_K_M | Details → | |
| 11 | Qwen3-Coder-30B-A3B-Instruct30BQwen (Alibaba) Performance benchmarkTimebench 3 - Kombi (Prefill + Generation) | 58,29 tok/s TG Prefill 3.005 · TTFT 4.672 ms | 5× | NVIDIA GeForce RTX 4090AMD Ryzen 3 3100 4-Core Processor · 61 GB RAM | llama.cppopenclaw_cliQ4_K_M | Details → | |
| 12 | Meta-Llama-3.1-8B-Instruct8BMeta Performance benchmarkTimebench 3 - Kombi (Prefill + Generation) | 53,50 tok/s TG Prefill 7.717 · TTFT 319 ms | 1× | NVIDIA GeForce RTX 4090AMD Ryzen 3 3100 4-Core Processor · 61 GB RAM | vLLMopenclaw_cliAWQ | Details → | |
| 13 | Qwen3-0.6B0.6BQwen (Alibaba) Performance benchmarkTimebench 3 - Kombi (Prefill + Generation) | 52,80 tok/s TG Prefill 20.717 · TTFT 121 ms | 1× | NVIDIA GeForce RTX 4090AMD Ryzen 3 3100 4-Core Processor · 61 GB RAM | vLLMopenclaw_cliFP8 | Details → | |
| 14 | Ministral-3-14B-Reasoning-251214BMistral AI Performance benchmarkTimebench 3 - Kombi (Prefill + Generation) | 50,31 tok/s TG Prefill 5.478 · TTFT 413 ms | 1× | NVIDIA GeForce RTX 4090AMD Ryzen 3 3100 4-Core Processor · 61 GB RAM | llama.cppopenclaw_cliQ8_0 | Details → | |
| 15 | Devstral-Small-2-24B-Instruct-251224BMistral AI Performance benchmarkTimebench 3 - Kombi (Prefill + Generation) | 49,20 tok/s TG Prefill 3.616 · TTFT 593 ms | 1× | NVIDIA GeForce RTX 4090AMD Ryzen 3 3100 4-Core Processor · 61 GB RAM | llama.cppopenclaw_cliQ4_K_M | Details → | |
| 16 | Devstral-Small-2-24B-Instruct-251224BMistral AI Performance benchmarkTimebench 3 - Kombi (Prefill + Generation) | 49,02 tok/s TG Prefill 3.516 · TTFT 610 ms | 1× | NVIDIA GeForce RTX 4090AMD Ryzen 3 3100 4-Core Processor · 61 GB RAM | llama.cppopenclaw_cliQ4_K_M | Details → | |
| 17 | Qwen3-Coder-30B-A3B-Instruct30BQwen (Alibaba) Performance benchmarkTimebench 3 - Kombi (Prefill + Generation) | 42,25 tok/s TG Prefill 1.875 · TTFT 1.298 ms | 1× | NVIDIA GeForce RTX 4090AMD Ryzen 3 3100 4-Core Processor · 61 GB RAM | llama.cppopenclaw_cliQ4_K_M | Details → |