§ AI Wiki / Glossary
One-line definitions, the AI dictionary.
§ Search this category
Search the Wiki →A closed-source commercial image generator known for its aesthetic quality.
Mistral AI's first open-weight language model, known for strong performance at a small size.
Paris-based AI lab and the flagship of the European open-weight LLM wave.
Mistral AI's open-weight Mixture-of-Experts language model family with strong performance per active parameter.
A large language model that also processes modalities like image, audio, or video.
A broad multiple-choice benchmark that tests knowledge and reasoning across 57 subjects.
An image-synthesis technique that represents a scene's 3D radiance field with a neural network.
Empirical relationships describing how model performance changes with parameters, data, and compute.
Systematic skews in a model's outputs that favor certain groups or viewpoints, usually inherited from training data or design choices.
The San Francisco–based AI company behind the GPT series and the ChatGPT product.
OpenAI's 2024 release introducing inference-time long thinking — the first major reasoning model.
The next-generation member of OpenAI's o-series reasoning models, focused on deeper reasoning and agent use.
Capabilities that appear in models only above a certain scale and are absent in smaller variants.
A machine learning paradigm in which the model generates its own training signal from unlabeled data.
A machine learning paradigm in which an agent learns a policy by interacting with an environment and optimizing a reward signal.
Microsoft's family of small LLMs trained heavily on high-quality data.
An attack where an adversary tries to override an LLM's instructions via untrusted external text.
A feature that caches large recurring prompt prefixes for major cost and latency savings.
Alibaba's 2024 generation of its open-weight, multilingual LLM family.
The practice of probing an AI system's limits and weaknesses with adversarial methods.
When a model declines to fulfill a request on safety or policy grounds.
A classic summarization metric based on n-gram and sequence overlap.
A New York–based company focused on creative industries that productized AI video generation.
Training data generated by another model rather than (or in addition to) real-world data.