§ AI Wiki / Glossary
One-line definitions, the AI dictionary.
§ Search this category
Search the Wiki →The January 2025 inflection point when DeepSeek R1's launch showed open-weight models catching up with frontier closed models.
The vision of national AI infrastructure where countries retain control over their language, data, and values.
Anthropic's user-visible long-thinking mode introduced for Claude models.
An AI-powered code completion and chat tool, launched in 2021 through a collaboration between GitHub and OpenAI.
An AI hardware company offering very high tokens-per-second inference through its LPU chips.
The company behind the Transformers library and the Model Hub that has become the central distribution layer for open AI models.
Autonomous AI agents that can read files, run commands, and produce code in a developer environment.
Microsoft's Copilot product family and internal AI organization, central to the strategic partnership with OpenAI.
A model gateway that unifies access to many LLM providers behind a single API.
An AI infrastructure platform that lets you run open-weight models behind a single API.
An approach where a model spends more compute at inference time, not just during training, to lift performance.
The competitive period that emerged in 2024 around inference providers competing on tokens per second (TPS).
When a model performs extensive internal reasoning across many tokens before producing its final answer.
An AI-first code editor developed by Codeium.
The AI company founded in 2023 by Elon Musk, the team behind the Grok product.
The community movement of developers and end users running LLMs locally on their own machines.
A Chinese AI lab behind the GLM family of language models.
A release model where only the trained weights are published, not the training code or dataset.
A large language model whose weights, code, or training material are publicly released.
The San Francisco–based AI company behind the Claude model family, founded with an explicit safety focus.
The schema describing a tool's name, description, and parameter types so the model can call it correctly.
A hypothetical AI that decisively surpasses human intelligence across virtually all cognitive tasks.
When a model refuses harmless or reasonable requests it should have answered.
A classic machine-translation metric based on n-gram overlap with reference translations.