COLLECTIONS.md

# Open Model Weights — two Hugging Face Collections

The ten linked publisher repositories were verified as existing public model pages on 12 October 2026. These lists are editorial starting points, **not rankings, independent benchmarks or endorsements**. Do not upload/copy publishers' weights.

## Collection 1 — Open-Weight Models Worth Exploring

**Description:** A cross-publisher starting point for open-weight AI models. Review the source cards, licenses and deployment requirements.

Add items in this order; paste the notes individually if desired:

1. [Qwen/Qwen3-32B](https://huggingface.co/Qwen/Qwen3-32B) — A Qwen3 model for comparing documented reasoning modes, language coverage and deployment trade-offs.
2. [deepseek-ai/DeepSeek-R1](https://huggingface.co/deepseek-ai/DeepSeek-R1) — A reasoning-focused release; assess model card details and hardware requirements.
3. [openai/gpt-oss-120b](https://huggingface.co/openai/gpt-oss-120b) — An OpenAI open-weight release; review formats, deployment options and documentation.
4. [mistralai/Mistral-Small-3.1-24B-Instruct-2503](https://huggingface.co/mistralai/Mistral-Small-3.1-24B-Instruct-2503) — Examine the instruction model's published configuration and modality requirements.
5. [google/gemma-3-27b-it](https://huggingface.co/google/gemma-3-27b-it) — A Gemma family release; review Google's specific license terms.

**Additional note:** Models are not ordered by quality. Explore source-linked comparisons at https://openmodelweights.com/compare/ .

## Collection 2 — Open-Weight Models for Local Deployment

**Description:** Compact model candidates for evaluating local and self-hosted AI. Real hardware fit depends on format, quantization, context and runtime.

Add items in this order; paste the notes individually if desired:

1. [Qwen/Qwen3-0.6B](https://huggingface.co/Qwen/Qwen3-0.6B) — Compact Qwen3 variant; test compatibility and actual device fit.
2. [Qwen/Qwen3-4B](https://huggingface.co/Qwen/Qwen3-4B) — Smaller Qwen3 release; inspect context and serving options.
3. [HuggingFaceTB/SmolLM3-3B](https://huggingface.co/HuggingFaceTB/SmolLM3-3B) — Compact Hugging Face model to evaluate for constrained inference.
4. [microsoft/Phi-4-mini-instruct](https://huggingface.co/microsoft/Phi-4-mini-instruct) — Compact instruction model; check license and intended deployment scope.
5. [google/gemma-3-4b-it](https://huggingface.co/google/gemma-3-4b-it) — Smaller Gemma release; check usage terms and memory requirements.

**Additional note:** Collection membership does not prove local performance or compatibility. Review hardware considerations at https://openmodelweights.com/hardware/ .

## Maintaining the Collections

Keep the first three entries informative: Hugging Face previews display up to three items per Collection. Verify source cards and publisher license terms before editing descriptions. Prefer factual notes over claims such as “best model” or “runs on all laptops.”