skills.rest
cloudrift-ai
cloudrift-aiOFFICIAL
Explore all skills by this author →
GitHub

discover-new-models

Rank newly released open-weight LLMs by demand and VRAM compatibility.

75|8|Updated Aug 2, 2025
Search Tags:#llm#openrouter#model discovery#huggingface#gpu benchmarking#open-weight models#vram fit
One-click install
npx skills add https://github.com/cloudrift-ai/emmy --skill discover-new-models
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: discover-new-models
Source: https://github.com/cloudrift-ai/emmy/tree/main/.claude/skills/discover-new-models
Command: npx skills add https://github.com/cloudrift-ai/emmy --skill discover-new-models

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) components.

What problem does it solve?

This Skill removes the manual effort of scouting for newly released, high-demand open-weight LLMs that are compatible with your team's GPU fleet and worth the compute investment to benchmark.

Core Features & Use Cases

  • Curated Candidate Sourcing: Pulls vetted, newly released open-weight models from public OpenRouter and HuggingFace endpoints, automatically excluding models already supported by emmy.
  • Demand-Aware Ranking: Ranks candidates by real-time popularity signals (HuggingFace trending score, LMArena Elo, releasing lab reputation) to prioritize models with genuine mindshare and proven quality.
  • Hardware Fit Mapping: Calculates VRAM requirements for available quantizations and maps each shortlisted model to compatible GPU configurations in your fleet.
  • Use Case: If your team wants to expand your emmy benchmark suite with the latest high-performance open models that run on your H200 and B200 GPUs, this Skill delivers a ready-to-use ranked shortlist pre-vetted for hardware compatibility and demand.

Quick Start

Use the discover-new-models skill to generate a ranked shortlist of new open-weight models compatible with your H100, H200, and B200 GPUs that are worth benchmarking next.

Frequently Asked Questions about discover-new-models

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I find new open-weight LLMs to benchmark on my GPU fleet?▼

To find new open-weight LLMs for benchmarking, you can automatically source trending models from HuggingFace and OpenRouter, filtering by popularity signals and VRAM compatibility for your specific GPUs. This produces a ranked, hardware-bucketed shortlist of vetted candidates.

How do I rank trending HuggingFace models by VRAM fit for different quantizations?▼

You rank trending HuggingFace models by calculating VRAM requirements for available quantizations and mapping each model to compatible GPU configurations. This filters candidates based on real hardware fit constraints for your available fleet.

Can I filter open-weight models by LMArena Elo and HuggingFace popularity before benchmarking?▼

Yes, you can filter open-weight models by combining LMArena Elo scores, HuggingFace trending metrics, and releasing lab reputation. This demand-aware ranking prioritizes models with proven quality and genuine mindshare for your benchmarking pipeline.

What is the best way to identify unsupported LLMs for an emmy benchmark suite?▼

The best way to identify unsupported LLMs for an emmy benchmark suite is to pull vetted models from public endpoints and automatically exclude any already supported models. This leaves a curated shortlist of newly released candidates.

Does this model discovery approach work with H100, H200, and B200 GPU configurations?▼

Yes, this model discovery approach works with H100, H200, and B200 GPU configurations by mapping shortlisted open-weight models to compatible hardware buckets. It calculates VRAM requirements to ensure each candidate fits your available fleet.

Related Skills

Explore More Software Engineering Skills→
JJKMatthewJJKMatthew

github-monitor

Monitor and aggregate StayAtHome GitHub trending data with Node.js and Express.

Community
Intermediate
scottgl9scottgl9
12

sglang-auto-benchmark

Automate repeatable LLM benchmarking across hardware and model configurations.

Community
Advanced
GeargrindadminGeargrindadmin

llmfit-advisor

Analyze local CPU, RAM, and GPU/VRAM to recommend LLM models with quantization and fit scores.

Community
Intermediate
lazyFrogLOLlazyFrogLOL
127

mteb-leaderboard

Locate and verify MTEB leaderboard standings for embedding models.

Community
Intermediate