EcoCompute Should-I-Quantize Estimator

io.github.hongping-zhPublished by GitHub account hongping-zhv1.0.0Updated Oct 24, 2026

Energy impact of LLM quantization, calibrated to measured GPU data. Every answer states its basis.

VerifiedStreamable HTTPWeb executableCloud & InfrastructureData & AnalyticsAI & ML

Overview

AI-generated overview

Remote MCP server that lets an assistant estimate the task-level energy change of LLM weight-only quantization (NF4/INT8) versus FP16 from measured GPU data.

What you can do
Estimate the task-level energy change of NF4 or INT8 weight-only quantization versus FP16 for a model size on a GPU architecture class, with basis, confidence and evidence boundaries. Look up the model-size crossover per architecture where quantization starts saving energy instead of costing it. List what is measured, borrowed or estimated, and where the coverage gaps are, before trusting an estimate at the edges. Search measured anchors by model name, GPU name, architecture or size, with energy change and run counts.
When to use it
Use it when comparing the task-level energy of NF4 or INT8 weight-only quantization against FP16 on a given GPU architecture class, or when checking whether a model size is past the point where quantization saves energy. Also useful for confirming data coverage before relying on an estimate.
Requirements
Remote endpoint; no local package to install. No authentication, environment variables or headers are declared. Network access is required.
Before you install
Numbers are GPU package power, not wall power, so they are not total electricity use. Combinations outside coverage are reported as gaps rather than extrapolated; distinguish measured from interpolated or estimated results by the stated basis and confidence. Contributing a measurement means going to an external code repository.

Installation

In SourceWeft

  1. Open EcoCompute Should-I-Quantize Estimator in the dashboard and add it to a workspace.
  2. Enable the server for the chats that should use its tools.

Web executable via Streamable HTTP. Remote servers run from the web runtime once configured in a workspace.

Use with an agent

Paste this prompt into Claude Code, Codex, Cursor or another agent — or into SourceWeft chat. The agent reads this server's install guide, shows you its publisher, endpoint and sign-in, and adds it to your client once you agree. Keys never go into a command.

Read https://sourceweft.com/mcp/io.github.hongping-zh%2Fecocompute-estimator/install.md and connect the MCP server it describes to my client. Before changing any config, show me its publisher, endpoint and sign-in method and wait for my OK. Never put an API key in a command or in a file I would commit; ask me where it should come from.

Read the install guide the agent follows

Other MCP clients

Add this to your client's mcpServers config.

{
  "mcpServers": {
    "ecocompute-should-i-quantize-estimator": {
      "type": "http",
      "url": "https://ecocompute-estimator.zhanghongping1982.workers.dev/mcp"
    }
  }
}

README

This repository has no README.

Tools

4

estimate_quantization_energy

Estimate the task-level energy change (ΔE% vs FP16) of weight-only quantization (NF4 or INT8) for a model of a given size on a GPU architecture class, from measured-anchored curves. The answer carries basis (measured/interpolated/estimated), confidence and evidence boundaries.

Parametersqctxarchbatchparams_bprecision

get_quantization_crossover

Get the model-size crossover for NF4/INT8 per architecture: below it quantization costs energy, above it saves. Crossovers belong to a card + software stack, not an architecture — the answer says so.

Parametersarch

list_data_coverage

List what is actually measured (GPU architecture × precision × size range, NVML direct), what is borrowed/estimated, and where the gaps are. Use this before trusting an estimate at the edges of coverage, and to answer 'is there data for my GPU?' honestly.

Takes no parameters.

search_measurements

Search the measured anchors by model name, GPU name, architecture or size. Returns ΔE% vs FP16 per anchor with run counts — the raw evidence behind the curves.

Parametersqueryrequired

Version history

1
  1. v1.0.0LatestOct 24, 2026

Score

75% of the applicable checks met

How complete this listing is, computed from the registry entry, the endpoint and public sources. Not a rating of the server and not a statement that it is safe or working today: reachability is a dated check, and no installs or reviews go into the score.

  • Reached in the last 30 daysRequired25 pts
  • Auth declaredRequired10 pts
  • Tools listedRequired15 pts
  • DescriptionRequired15 pts
  • Logo10 pts
  • Documentation15 pts

    Declare a GitHub repository with a README, or serve resource_documentation in the OAuth protected-resource metadata.

  • License10 pts

    Declare a license in the repository or the registry entry.