Compute:Arena collects community benchmarks for local models across chips, runtimes and quantization settings, including decode and prefill throughput tied to each submitted test setup.
Features
- Community benchmark data
- Chip and quantization comparisons
- Throughput metrics
Use Cases
- Evaluating local inference hardware
- Comparing model configurations
FAQ
Compute:Arena collects community benchmarks for local models across chips, runtimes and quantization settings, including decode and prefill throughput tied to each submitted test setup. Core capabilities include: Community benchmark data, Chip and quantization comparisons, Throughput metrics.
Common scenarios include: Evaluating local inference hardware, Comparing model configurations.
