gemma-2-9b-it-4bit
7 workload results across 3 hardware configurations.
Fastest local config
152.7 decode tok/s
on RTX 5090 (32GB) + AMD Ryzen 7 9850X3D 8-Core Processor (8c) + 30GB via llama.cpp. see full run
For a standard 4-bit quantization using MLX, here are measured results for short-chat generation.
On MLX 4-bit, gemma-2-9b-it is fastest on the M3 Ultra (60-core GPU) at 89.45 tok/s, 307% faster than the M3 Pro (18-core GPU) (21.97 tok/s).
| Card | Decode tok/s | TTFT (118-token prompt) | Run |
|---|---|---|---|
| M3 Ultra (60-core GPU) | 89.45 (single run) | 691 ms | r_iz137eqvuzy |
| M3 Pro (18-core GPU) | 21.97 (single run) | 1151 ms | r_q7t7b7dcuz5 |
gemma-2-9b-it, 4-bit, MLX, chat-short, batch 1. Only cards with their standard memory are listed; CPU-only hosts are excluded.
Local runs (7 workload results)
Runs from contributors' own machines via MLX, llama.cpp, vLLM, exllamav2, or ollama. Signed on the submitter's hardware.
RTX 5090 (32GB) + AMD Ryzen 7 9850X3D 8-Core Processor (8c) + 30GB
| Workload | Backend | Quant | decode tok/s | prefill tok/s | TTFT | Run |
|---|---|---|---|---|---|---|
| chat-short | llama.cpp | - | 69.45tok/s | no data | 325ms | r_1_xl4zb5-xj |
| chat-long | llama.cpp | - | 69.42tok/s | no data | 1,044ms | r_1_xl4zb5-xj |
| concurrent-decode | llama.cpp | - | 68.57tok/s | no data | no data | r_1_xl4zb5-xj |
| agent-trace | llama.cpp | - | 68.46tok/s | 4,816.8tok/s | 420ms | r_1_xl4zb5-xj |
| chat-short | llama.cpp | - | 152.7tok/s | no data | 58.9ms | r__b_bzmmab_8 |
M3 Pro (18-core GPU) + 36GB unified
| Workload | Backend | Quant | decode tok/s | prefill tok/s | TTFT | Run |
|---|---|---|---|---|---|---|
| chat-short | mlx@0.31.3 | - | 21.97tok/s | 102.5tok/s | 1,151ms | r_q7t7b7dcuz5 |
M3 Ultra (60-core GPU) + 96GB unified
| Workload | Backend | Quant | decode tok/s | prefill tok/s | TTFT | Run |
|---|---|---|---|---|---|---|
| chat-short | mlx@0.31.3 | - | 89.45tok/s | 170.7tok/s | 691ms | r_iz137eqvuzy |
gemma-2-9b-it-4bit on hardware
Request a gemma-2-9b-it-4bit run
No measured run yet for these pairs. Vote for the one you want; the most-requested runs go first. See all requests.
We never sell or share your email. Privacy