All servers · AI models & agents
EcoCompute Should-I-Quantize Estimator
io.github.hongping-zh/ecocompute-estimator · v1.0.0
New (7 days)
Measured by MCP Station
| Last check | Reachable |
|---|---|
| Checked at | — |
| Check success 7 / 30 days | — |
| First-try success 30 days | — |
| Median response | — |
| Lookups 30 days | — |
| Authentication | No auth |
| MCP version | 2025-06-18 |
| Rank score | — |
| Declared risk | Not declared (tools carry no read/destructive hints) |
| Capabilities declared | tools |
| Tool definitions | 4 tools · with description 100% · input schema is an object 100% · read/destructive hints 0% |
Description written by the publisher, not verified
Energy impact of LLM quantization, calibrated to measured GPU data. Every answer states its basis.
Connect
streamable-http · https://ecocompute-estimator.zhanghongping1982.workers.dev/mcp
claude mcp add --transport http ecocompute-estimator https://ecocompute-estimator.zhanghongping1982.workers.dev/mcp{
"mcpServers": {
"ecocompute-estimator": {
"url": "https://ecocompute-estimator.zhanghongping1982.workers.dev/mcp"
}
}
}Tools as reported by the server
estimate_quantization_energyEstimate 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.get_quantization_crossoverGet 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.list_data_coverageList 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.search_measurementsSearch 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.
Repository · Website · JSON · Text for agents · Registry updated 2026-10-24