0
Quantized builds
39
Benchmark records
Quantized builds
Community and official quantized releases; "linked records" counts only benchmarks tied to that exact release.
No quantized builds cataloged yet.
Benchmark data
39 configurations
| Hardware | Framework | Quant | Decode | VRAM used | Samples | Evidence |
|---|---|---|---|---|---|---|
| AMD Ryzen AI Max+ 395 | llama.cpp | IQ2_XS | 92.6 tok/s | 28.31 GB | 1 | L2 Cross-framework verified |
| AMD Ryzen AI Max+ 395 | llama.cpp | IQ3_XS | 81.1 tok/s | 35.95 GB | 1 | L2 Cross-framework verified |
| AMD Ryzen AI Max+ 395 | Ollama | IQ3_XS | 79.4 tok/s | 35.95 GB | 1 | L2 Cross-framework verified |
| Apple M4 Pro | vLLM | IQ3_XS | 87.6 tok/s | 35.95 GB | 1 | L2 Cross-framework verified |
| Apple M4 Pro | vLLM | Q4_K_S | 72.5 tok/s | 44.69 GB | 1 | L2 Cross-framework verified |
| Apple M4 Pro | Ollama | Q5_K_M | 56.2 tok/s | 52.88 GB | 1 | L2 Cross-framework verified |
| Apple M4 Pro | Ollama | Q5_K_S | 59.3 tok/s | 51.24 GB | 1 | L2 Cross-framework verified |
| Apple M5 Pro | llama.cpp | Q8_0 | 55.3 tok/s | 60.8 GB | 1 | L2 Cross-framework verified |
| Apple M5 Pro | Strata | Q6_K | 98.9 tok/s | 56.7 GB | 1 | L2 Cross-framework verified |
| Apple M5 Max | vLLM | Q3_K_S | 174.4 tok/s | 37.59 GB | 1 | L2 Cross-framework verified |
| Apple M5 Max | vLLM | Q4_K_S | 136.4 tok/s | 44.69 GB | 1 | L2 Cross-framework verified |
| Apple M5 Max | Strata | Q2_K | 229 tok/s | 32.13 GB | 1 | L2 Cross-framework verified |
| Apple M5 Max | Strata | Q8_0 | 122.6 tok/s | 64.89 GB | 1 | L2 Cross-framework verified |
| NVIDIA DGX Spark | llama.cpp | IQ3_S | 112.4 tok/s | 38.14 GB | 1 | L2 Cross-framework verified |
| NVIDIA DGX Spark | llama.cpp | Q3_K_L | 105.6 tok/s | 41.41 GB | 1 | L2 Cross-framework verified |
| NVIDIA DGX Spark | Ollama | Q3_K_L | 104.2 tok/s | 41.41 GB | 1 | L2 Cross-framework verified |
| NVIDIA DGX Spark | Ollama | Q3_K_M | 113.1 tok/s | 39.23 GB | 1 | L2 Cross-framework verified |
| NVIDIA Jetson AGX Orin 64GB | vLLM | IQ4_XS | 42.8 tok/s | 42.5 GB | 1 | L2 Cross-framework verified |
| NVIDIA Jetson AGX Orin 64GB | vLLM | Q5_K_S | 34.1 tok/s | 51.24 GB | 1 | L2 Cross-framework verified |
| NVIDIA Jetson AGX Orin 64GB | Ollama | Q8_0 | 24.7 tok/s | 60.8 GB | 1 | L2 Cross-framework verified |
| NVIDIA RTX 5070 Ti | vLLM | IQ2_XS | 86.5 tok/s | 9.3 GB | 1 | L0 Self-reported |
| NVIDIA RTX 5070 Ti | vLLM | Q5_K_M | 39.5 tok/s | 16 GB | 1 | L0 Self-reported |
| NVIDIA RTX 5070 Ti | vLLM | Q4_K_M | 48.2 tok/s | 16 GB | 1 | L0 Self-reported |
| NVIDIA RTX 5070 Ti | llama.cpp | IQ2_XS | 79.4 tok/s | 9.3 GB | 1 | L0 Self-reported |
| NVIDIA RTX 5070 Ti | llama.cpp | Q5_K_M | 36.2 tok/s | 16 GB | 1 | L0 Self-reported |
| NVIDIA RTX 5070 Ti | llama.cpp | Q4_K_M | 44.2 tok/s | 16 GB | 1 | L0 Self-reported |
| NVIDIA RTX 5090 | vLLM | IQ2_XS | 173.1 tok/s | 9.3 GB | 1 | L0 Self-reported |
| NVIDIA RTX 5090 | vLLM | Q5_K_M | 78.9 tok/s | 19.8 GB | 1 | L0 Self-reported |
| NVIDIA RTX 5090 | vLLM | Q4_K_M | 96.4 tok/s | 16.3 GB | 1 | L0 Self-reported |
| NVIDIA RTX 5090 | llama.cpp | IQ2_XS | 158.8 tok/s | 9.3 GB | 1 | L0 Self-reported |
| NVIDIA RTX 5090 | llama.cpp | Q5_K_M | 72.4 tok/s | 19.8 GB | 1 | L0 Self-reported |
| NVIDIA RTX 5090 | llama.cpp | Q4_K_M | 88.5 tok/s | 16.3 GB | 1 | L0 Self-reported |
| NVIDIA RTX 4090 | vLLM | IQ2_XS | 97.4 tok/s | 9.3 GB | 1 | L0 Self-reported |
| NVIDIA RTX 4090 | vLLM | Q5_K_M | 44.4 tok/s | 19.8 GB | 1 | L0 Self-reported |
| NVIDIA RTX 4090 | vLLM | Q4_K_M | 54.2 tok/s | 16.3 GB | 1 | L0 Self-reported |
| NVIDIA RTX 4090 | llama.cpp | IQ2_XS | 89.3 tok/s | 9.3 GB | 1 | L0 Self-reported |
| NVIDIA RTX 4090 | llama.cpp | Q5_K_M | 40.7 tok/s | 19.8 GB | 1 | L0 Self-reported |
| NVIDIA RTX 4090 | llama.cpp | Q4_K_M | 49.8 tok/s | 16.3 GB | 1 | L0 Self-reported |
| Apple M2 Ultra | llama.cpp | Q4_K_M | 79.1 tok/s | — | 1 | L1 Reproduced |
39 of these records are not linked to a specific quantized build; the full set is on the search page.
Community discussion
No discussions yet — start the first one.
Start a discussion