🏆
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 | Phi-3.5-mini-instruct3.8Bmicrosoft Performance benchmarkTimebench 3 - Kombi (Prefill + Generation) | 272,93 tok/s TG Prefill 691 · TTFT 38.379 ms | 10× | 2x AMD Radeon RX 7900 XTX i4x Intel(R) Xeon(R) CPU E5-4657L v2 @ 2.40GHz · 504 GB RAM | llama.cppgodclawQ4_K_M | Details → | |
| 2 | Phi-3.5-mini-instruct3.8Bmicrosoft Performance benchmarkTimebench 3 - Kombi (Prefill + Generation) | 172,34 tok/s TG Prefill 5.226 · TTFT 11.908 ms | 5× | 2x AMD Radeon RX 7900 XTX i4x Intel(R) Xeon(R) CPU E5-4657L v2 @ 2.40GHz · 504 GB RAM | llama.cppgodclawQ4_K_M | Details → | |
| 3 | Phi-3.5-mini-instruct3.8Bmicrosoft Performance benchmarkTimebench 3 - Kombi (Prefill + Generation) | 99,48 tok/s TG Prefill 4.127 · TTFT 671 ms | 1× | 2x AMD Radeon RX 7900 XTX i4x Intel(R) Xeon(R) CPU E5-4657L v2 @ 2.40GHz · 504 GB RAM | llama.cppgodclawQ4_K_M | Details → |