liveBook · Manning

Open-source model

As explained in Generative AI in Action

An open-source large language model is a language model whose source artifacts and permitted use are made available under an open-source license.

An open-source large language model makes its source artifacts and permitted uses available under an open-source license. Enterprises weigh it against commercial models when selecting technology for a specific use case, balancing customization, integration, cost, security, and regulatory requirements. section 1.7 section 2.9

Evaluate an open-source model across these practical dimensions:

  • Compare capability with commercial models; smaller foundational models increasingly target similar quality levels. section 2.7
  • Check the license before deployment, including whether commercial use, extension, and modification are permitted. section 1.7 section 2.7.2
  • Plan the integration path: some models can be hosted on Hugging Face and called through a Visual Studio Code extension or custom endpoint. section 5.3.3
  • Treat enterprise governance as a separate requirement; commercial platforms may provide privacy controls, customer-managed keys, encryption, private networking, regional availability, and content filtering. section 2.7.2

Dolly v2 as a licensing example

Databricks labels Dolly v2 a truly open instruction-tuned LLM. Its CCA 3.0 license allows use, extension, and modification, including commercial use; the model has 12 billion parameters. section 2.7.2

Table 2.2 Open source LLMs
Company
Open source LLM
Parameter size
Meta
Llama LLM is one of the models that has inspired many other OSS models. It comes in many sizes (7B, 13B, 33B, and 65B), and while smaller than GPT-3, it can be matched across many tasks. Meta shared the models with researchers (and they were also leaked separately online), inspiring many others to use that as a jumping-off point.
Various (7B–65B)
Databricks
Databricks recently released v2 of Dolly, which they label the “world’s first truly open instruction-tuned LLM.” It is released under the CCA 3.0 license, allowing anyone to use, extend, and modify it, including for commercial purposes.
12B
Alpaca
Standford University’s Alpaca, an instruction model based on Llama, claims to match the GPT-3.5 Turbo performance in some tasks.
7B
FreedomGPT
This is an OSS conversational agent based on Alpaca. They claim to offer 100% uncensored and private conversations.
Not disclosed
Vicuna
Academic researchers from multiple institutions (UC Berkeley, CMU, Stanford, UC San Diego, and MBZUAI) released Vicuna, a fine-tuned version of Llama that matches the GPT4 performance across many tasks.
13B
Koala
Berkley AI Research released Koala, a fine-tuned version of Llama using internet dialogues.
13B
ChatLLaMa
Technically, it’s not a model but tooling for models. Nebuly AI released ChatLLaMa, a library that can create ChatGPT-like conversational assistance using your data.
7B
ColossalChat
UC Berkeley’s ColossalAI project released ColossalChat, a ChatGPT-like model that includes complete RLHF pipelines based on Llama.
7B
Falcon
The Technology Innovation Institute (TII) in the United Arab Emirates released a family of LLMs called the Falcon LLM model. At the time, Falcon was the biggest OSS LLM ever released and was at the top of the OSS LLM Leaderboard. More recently, a more powerful 180B parameters model is again at the top of the leaderboard.
Various (1B–180B)
Mistral
Mistral AI, a French startup, has developed a range of models. Some are open source models licensed under Apache 2.0, a permissive license allowing unrestricted use in any context. As mentioned in the previous chapter, they also have commercial models.
Various (7B – 141B)

Browse a comparison of open-source LLM examples, organizations, parameter sizes, licenses, and claimed capabilities. section 2.7.2

How Generative AI in Action covers it

How deep it goes
Defines the category and compares examples, licenses, capabilities, and integration paths without detailing training or inference internals.
Examples
Uses Llama, Dolly, Alpaca, FreedomGPT, StableCode, Hugging Face hosting, and Visual Studio Code integration as concrete examples.
What it assumes
Assumes basic familiarity with LLMs and foundational models while positioning open-source models alongside commercial platforms.
What you can do after
Evaluate/understand model choices by matching capabilities, licenses, integration options, and enterprise requirements to a use case.
How it connects
Connects open-source models with commercial models, foundational models, instruction tuning, fine-tuning, small language models, and enterprise governance.
Cover of Generative AI in Action
liveBookmanning.com
Read Generative AI in Action on liveBook Get the book at manning.com

Also explained in

Compare how 5 books explain open-source model →

Related concepts

This explainer was generated by Manning's knowledge graph from the text of Generative AI in Action. The section links point to the parts of the book it draws on.