# ToshLLM benchmarks

> Public measurements of local models on Intel Macs with AMD GPUs. The site groups comparable runs by GPU and model and keeps individual signed evidence accessible.

Benchmark measurements are observations from particular hardware and software configurations. Compare prompt processing and token generation separately. A signed community run confirms the app recorded and submitted the payload; only explicitly verified runs are ToshLLM Lab measurements. Review model name, quantization, GPU, memory, macOS, app version and configuration before quoting a speed.

## Results and methodology

- [Benchmark index](https://toshllm.com/benchmarks): Filterable public results and grouped GPU/model comparisons.
- [Benchmark guide](https://toshllm.com/docs/benchmarks): How results are measured and what they mean.
- [Submission process](https://toshllm.com/contribute): App-only review, signing and sharing flow.
- [Public JSON API](https://toshllm.com/api/v1/benchmarks?limit=100): Paginated machine-readable results.

## Optional

- [Documentation](https://toshllm.com/docs): Product context and compatibility.