VynarisEarly beta Kimi K3Get your API key

Glossary · Category

Ecosystem & products

65 plain-English definitions.

AGI (artificial general intelligence)
A hypothetical future AI system capable of matching or exceeding human performance across essentially all cognitive tasks.
AI app builder
A no-code or low-code platform that lets users assemble LLM-powered applications without writing traditional software.
AI engineer
A relatively new job title/discipline focused on building applications on top of foundation models, distinct from traditional ML engineering.
Anyscale
The company behind the Ray distributed computing framework, offering a managed platform for scaling AI training and serving workloads.
API compatibility layer
A wrapper that makes one provider's API mimic another's (e.g. OpenAI-compatible endpoints) so existing client code works unchanged.
Baseten
A platform for deploying and serving custom machine learning models in production.
Braintrust
An evaluation and observability platform focused on testing and iterating on LLM application quality.
Chroma
A lightweight, developer-friendly open-source vector database popular for prototyping RAG applications.
Claude
Anthropic's family of LLMs, known for strong coding, reasoning, and long-context performance.
Cohere
An enterprise-focused AI lab known for retrieval and embedding models alongside its LLM offerings.
compound AI system
An application built from multiple interacting components (models, retrievers, tools, code) rather than a single monolithic model call.
context protocol ecosystem
The growing set of tools, servers, and integrations built around standards like MCP for connecting agents to external systems.
Databricks Mosaic AI
Databricks' platform for training, fine-tuning, and serving custom LLMs within a unified data and AI stack.
DeepSeek
A Chinese AI lab known for highly efficient open-weight reasoning and mixture-of-experts models that disrupted assumptions about training cost.
Fireworks AI
An inference platform focused on fast, cost-efficient hosting of open-weight and custom LLMs.
foundation model
A large, general-purpose model pretrained on broad data that can be adapted to many downstream tasks.
frontier lab
One of the small group of organizations (OpenAI, Anthropic, Google DeepMind, Meta, etc.) that develop and release the most capable AI models.
frontier model
The most capable class of LLMs available at a given time, typically from leading labs like OpenAI, Anthropic, or Google DeepMind.
Gemini
Google DeepMind's family of multimodal LLMs, integrated across Google's products and cloud platform.
GPT
OpenAI's family of generative pretrained transformer models, the namesake and one of the earliest widely deployed LLM product lines.
GPU cloud
A cloud provider specializing in renting out GPU compute capacity, often used for AI training and inference workloads.
Grok
xAI's family of LLMs, integrated into the X (Twitter) platform.
Groq
An AI hardware and inference company known for extremely fast token generation using its custom LPU chips.
Helicone
An open-source LLM observability platform for logging, monitoring, and analyzing API usage and cost.
Hugging Face
An open platform and hub for sharing, discovering, and deploying open-weight models, datasets, and ML tools.
Hugging Face Hub
The model and dataset repository where the open-source ML community publishes and shares pretrained models.
inference aggregator
A platform that provides unified access to many different hosted models and providers through one API and billing relationship.
inference provider
A company that offers hosted API access to run inference on LLMs, either their own or third-party open-weight models.
LangSmith
LangChain's platform for tracing, evaluating, and debugging LLM application runs in development and production.
LiteLLM
An open-source library and proxy that provides a unified API interface for calling dozens of different LLM providers.
Llama
Meta's family of open-weight LLMs, widely used as a base for self-hosted and fine-tuned deployments.
LLMOps
The set of practices and tooling for deploying, monitoring, and maintaining LLM applications in production, analogous to MLOps.
Milvus
An open-source vector database designed for large-scale similarity search workloads.
Mistral
A French AI lab known for efficient open-weight and commercial LLMs, including mixture-of-experts models.
MLOps
The discipline of operationalizing machine learning models, covering deployment, monitoring, and lifecycle management.
model API
The hosted endpoint through which developers send requests to a provider's model and receive generated responses.
model bakeoff
An informal or structured comparison where a team runs the same task across several candidate models to decide which to adopt in production.
model context ecosystem
The growing set of tools, servers, and integrations being built around agent-to-tool connectivity standards.
model leaderboard
A public ranking site that compares LLMs on benchmarks or head-to-head preference votes.
model marketplace
A platform where users can browse, compare, and access many different hosted or open-weight models in one place.
model weights leaderboard
A ranking specifically comparing open-weight models against each other on benchmarks, separate from closed-API leaderboards.
multimodal model
A model capable of processing and/or generating more than one data type, such as text, images, audio, or video together.
Open LLM Leaderboard
A Hugging Face-hosted leaderboard ranking open-weight models across standardized benchmarks.
OpenAI-compatible API
An API endpoint designed to accept the same request format as OpenAI's API, letting developers swap providers with minimal code changes.
OpenRouter
A unified API and marketplace that routes requests across many different LLM providers through a single interface.
Perplexity
An AI-powered answer engine that combines LLMs with real-time web search to produce cited, conversational search results.
pgvector
A PostgreSQL extension that adds vector similarity search capability directly inside a standard relational database.
Pinecone
A managed vector database service commonly used as the retrieval backend for RAG applications.
Portkey
An AI gateway product offering unified LLM access, caching, routing, and observability across providers.
prompt engineer
A role focused on designing and optimizing prompts to reliably get desired behavior from LLMs.
Qdrant
An open-source vector database optimized for high-performance similarity search in AI applications.
Qwen
Alibaba's family of open-weight LLMs, widely used in the open-source community for fine-tuning and self-hosting.
Replicate
A platform for running and deploying open-source machine learning models via a simple API.
RunPod
A GPU cloud provider offering on-demand and serverless GPU instances popular for AI inference and fine-tuning workloads.
Sentence Transformers
A widely used open-source Python library for computing sentence and text embeddings.
small language model (SLM)
A compact LLM (typically under a few billion parameters) designed to run cheaply and quickly, often on-device or at the edge.
speech-to-text model
A model that transcribes spoken audio into written text.
text embeddings inference (TEI)
Hugging Face's optimized serving toolkit specifically for running embedding models at high throughput.
text-to-speech (TTS) model
A model that converts written text into natural-sounding synthesized speech audio.
Together AI
An inference and fine-tuning platform specializing in hosting open-weight models at scale.
vision-language model (VLM)
A multimodal model that can process both images and text, answering questions about or reasoning over visual input.
Weaviate
An open-source vector database used for semantic search and RAG retrieval pipelines.
Weights & Biases (W&B)
A platform for tracking machine learning experiments, model training runs, and increasingly LLM evaluation workflows.
world model
A model trained to predict and simulate how an environment changes over time, an emerging research direction beyond pure language modeling.