Skip to content
modelread

Loading devices…

Remembered on this device. Multi-GPU: search “2x”, “8x”, “node”.

Changelog

2026-11-18

  • modelread.com; new website with model and device pages; MCP Registry entry com.modelread/modelread; Gradio MCP Space on Hugging Face.
  • MCP tools carry titles and read-only annotations; prompts and a hardware resource.
  • Quantized derivatives that are really fine-tunes rank below faithful quants; cross-model ranking accounts for quant quality loss.
  • Ask accepts a default device.

2026-11-17

  • Cloud instances (AWS, GCP, Azure, DigitalOcean, OCI, Hetzner, Lambda) with $/hour and cost per million tokens.
  • Phones and tablets calibrated from published measurements; Gemma 3n/4 per-layer embeddings and shared KV handled.
  • Tensor-parallel serving estimates for multi-GPU nodes; KLD-based quant quality.
Askwhat should I run?
Browsebest models per device
Devicesall hardware
Connect your agentMCP · API
How the numbers workcalibration