Item: NVIDIA/Model-Optimizer
Rating: 38.0

NVIDIA/Model-Optimizer

A unified library of SOTA model optimization techniques like quantization, distillation, pruning, neural architecture search, speculative decoding, etc. It compresses deep learning models for downstream deployment frameworks like TensorRT-LLM, TensorRT, vLLM, etc. to optimize inference speed.

[Apache-2.0]G10 · Libraries & Frameworks (SDKs)current state

Trust Score

38.0

Last full review 2026-09-26 · Trust Score decays −2 points per month without a fresh review (never below 20)

security cap applied — score capped at 45

100%evidence coverage 100%req_f84857852ddf

Six dimensions

Anchors 0 · 25 · 50 · 75 · 100

U Ministry Utility

“"A unified library of SOTA model optimization techniques like quantization, distillation, pruning, neural architecture search, speculative…”
source unavailable

M Maturity

“"latest tag 0.47.0 published 2026-09-23T18:55:03Z" (api.releases); "pushed_at": "2026-09-26T06:14:55Z" (api.repos); "fix(deps): declare…”
source unavailable

Movement

2026-09-25 → 2026-10-06 · 12 snapshots

38.8 → 40.0 (+1.2 over window)

ranked #5 of 11 in G10 (▼1 vs previous snapshot)

Plan your first step with NVIDIA/Model-Optimizer

NVIDIA/Model-Optimizer looks like a candidate for libraries & frameworks (sdks); unfetched: contributors — contributor breakdown not in pack; M score capped at 75 pending verification of >=2 active contributors.

server:Checking

Security: a review cap is active on this project — see the security note above.

Use it with

Selected: Gloo AI Studio

Tool names and logos belong to their owners; shown so you can pick your own client. Not affiliated.

Step 1 — Connect Refinry once (prompt only)

Gloo AI Studio does not document external MCP server setup. Use this prompt with the copied Refinry research brief in Studio.

https://studio.ai.gloo.com/

Connection prompt

I am choosing open-source software for my ministry using Refinry, a review site. Read Refinry’s plain-language agent feed — https://refinry.dev/docs.md — then help me evaluate NVIDIA/Model-Optimizer: summarize the Trust Score, the strongest and weakest dimensions, and any caution flags it cites. Give me one small first step I can complete this week, and tell me honestly if this looks like more than I can handle.

Connection didn't work?

Some clients need manual setup. Follow the plain-language steps on the Docs page.

Step 2 — Plan your first step

Tell the AI what you want to do; edit this to match your situation.

I want to…Starter prompt

I am choosing open-source software for my ministry using Refinry, a review site. Read Refinry’s public review of NVIDIA/Model-Optimizer (https://refinry.dev) and help me assess it for using NVIDIA/Model-Optimizer for libraries & frameworks (sdks) in my ministry. Summarize the Trust Score, the strongest and weakest dimensions, and any caution flags it cites, separating reviewed evidence from what still needs verification. Give me one small first step I can complete this week, and tell me honestly if this looks like more than I can handle.

Prefer a hand-off that works without a connection? Copy the research brief: every reviewed fact, all caution flags, and a feasibility-only first assignment.