LEADERBOARD
Fastest rigs, ranked by real benchmarks
Ranked from community-measured tokens/sec — not our estimates. Numbers from one or two reports are tagged prelim until a third confirms them. Add your benchmark →
Most tokens per dollar
Measured decode tok/s per $1,000 of device cost.
- 1.81.4tok/s per $1kNVIDIA RTX 4090 (24GB) · Llama 3.1 8B Q4_K_M · 130.2 tok/s · $1,599prelim
Fastest rig per model
Llama 3.1 8B
- 1.130.2tok/sNVIDIA RTX 4090 (24GB) · Q4_K_Mprelim
- 2.49.3tok/sMacBook Pro M2 Max (32GB) · Q4_K_Mprelim
3 benchmarks · 2 devices · 1 model · full benchmark table →