Question 1

How do I pick the best open LLM for a chatbot?

Accepted Answer

Focus on models with high context lengths (e.g., ChatGLM3-128k) and instruction-tuned versions like Llama 3 Instruct or Mistral Instruct. Check the license column for commercial use, and reference linked benchmarks for performance insights.

Question 2

What's the difference between Apache 2.0 and custom licenses for LLMs?

Accepted Answer

Apache 2.0 allows unrestricted commercial use and modification, while custom licenses like those for LLaMA or Qwen often restrict user scale (e.g., under 700M users) and prohibit using outputs to train competing models. Always review the specific terms in the checkpoint links.

Question 3

How to fine-tune an open LLM using the listed datasets?

Accepted Answer

Use datasets like OpenAssistant Conversations or databricks-dolly-15k from the instruction-tuning table, which are permissively licensed. Pair them with a base model like Pythia or MPT, and follow the linked papers or external guides for training code, as this directory doesn't provide implementation details.

Question 4

Is Llama 3 or Mistral better for code generation?

Accepted Answer

For code-specific tasks, prioritize models from the 'Open LLMs for code' table like Code Llama or StarCoder, which are optimized for programming. Between general models, Llama 3 has strong reasoning, but Mistral offers efficient inference with sliding windows; check external benchmarks for your use case.

Question 5

Where can I see up-to-date performance scores for these models?

Accepted Answer

The README links to several external leaderboards like the Hugging Face Open LLM Leaderboard and LMSYS Chatbot Arena. These provide holistic evaluations, but note that scores can vary by benchmark, so cross-reference multiple sources for reliability.

Question 6

How do I contribute or add a new model to the directory?

Accepted Answer

The README states 'Contributions welcome!'—submit a pull request on GitHub with the model details formatted in the table, including checkpoints, license, and context length. Ensure it meets the commercial license criteria and includes verifiable sources.

Open LLMs

What is Open LLMs?

Overview

Use Cases

Best For

Related Projects

Found a gem we're missing?

Not Ideal For

Pros & Cons

Pros

Cons

Frequently Asked Questions