Glossary
OSAID
The OSI's October 2024 definition of "open source AI," requiring not just weights but enough information about data, code, and architecture for third parties to reproduce the system.
The OSIgovernanceThe nonprofit that maintains the canonical Open Source Definition for software since 1998, and the OSAID definition for AI as of 2024. Open full entry ’s authoritative definition of open-source AI, released October 2024 after a two-year consultation. It demands four freedoms paralleling the Open Source Definition: freedom to use the system for any purpose, to study how it works, to modify it, and to share it (modified or not) with anyone.
For AI systems specifically, OSAID requires sufficient information about training data (not necessarily the data itself, which may be legally undistributable), the source code used to train and run the system, and the trained parameters. The data carve-out is contested: critics argue that without the actual training data a system is not reproducible, so the OSIgovernanceThe nonprofit that maintains the canonical Open Source Definition for software since 1998, and the OSAID definition for AI as of 2024. Open full entry ’s compromise admits systems that other definitions would exclude.
Under the strictest reading of OSAID, very few 2026 frontierweightsThe current capability envelope of AI, defined by the most capable models in deployment at any given time; an evolving label rather than a fixed threshold. Open full entry -class releases qualify. AI2’s OLMo family does. LlamaweightsMeta's open-weight model family, the most widely deployed open release through 2024 to 2026, released under the source-available Community License with an MAU cap and acceptable-use clause. Open full entry , MistralweightsA French open-weight model family from Mistral AI, released mostly under Apache 2.0 with strong performance per parameter and notable MoE variants (Mixtral, Mixtral 8x22B). Open full entry , QwenweightsAlibaba's open-weight model family, leading the multilingual and Chinese-language open-weight space, released under Apache 2.0 with sizes from 0.6B to 235B parameters. Open full entry , DeepSeekweightsA Chinese open-weight family known for the V3 MoE base model and the R1 reasoning model, both released under permissive licenses and unusually transparent in their training-cost reporting. Open full entry , and GemmaweightsGoogle's open-weight model family derived from Gemini research, with source-available licensing that includes an acceptable-use clause and license-revocation hook. Open full entry are mostly weight-only and do not. The OSAID’s practical effect is to provide a meaningful term for the open-AI community that vendor marketing has not yet diluted.