TOOLS
First identify whether a name is a model, application, coding agent, runtime, or local inference tool. Then inspect how it relates to context, tools, permissions, and result checking.
Verify names, organizations, product types, main interaction entry points, and concept associations; does not indicate feature rankings, version recommendations, or real-time capability conclusions.
Start with the task
One task may require a model, application, platform, or local runtime working together.
Then identify the layer
Models, applications, workspaces, platforms, agents, and local runtimes solve different problems.
Browse and compare
A set of foundational model endpoints provided by OpenAI. Applications call models through APIs; the models themselves do not directly include any application's interface or account permissions.
Suitable for distinguishing 'models' from 'applications that use models', and also explains why the same model might appear in different products.
A foundational model family provided by Meta, available for official APIs, open-weight deployment and local inference.
Suitable for explaining how open-weight models enter third-party applications, cloud services and local runtime environments.
Foundational models, assistant and agent-related technologies provided by Mistral AI.
Suitable for explaining how a model provider can simultaneously offer models, APIs, assistants and application products.
User-facing AI application offering conversation, files, tools, and memory features. Actual available capabilities depend on account, subscription, and product version.
Good for explaining how the application layer combines models, interfaces, context, tools, memory, and permissions, while users experience the application as a whole.
Microsoft's AI assistant entry point for individuals and organizations, connecting search, productivity apps, and work data based on products and accounts.
Good for explaining how the same assistant name may cover multiple products and permission environments, with enterprise data boundaries being particularly important.
AI research and understanding tool centered on user-provided datasets, suitable for Q&A, summarization, and organization around sources.
Good for explaining the relationship between retrieval augmentation, contextual sources, citation verification, and hallucination checking in data-centric applications.
ByteDance's user-facing AI assistant product, offering conversation, search, content processing, and related tool entry points.
Good for explaining how mass-market AI applications integrate models, search, files, tools, and content ecosystems.
Tencent's user-facing AI assistant product, combining search, content, and productivity capabilities within the Tencent ecosystem.
Good for explaining how AI applications are influenced by their platform ecosystem, account permissions, and available tool connections.
AI Q&A and research workstation focused on web retrieval and source citations.
Good for illustrating how search, source linking, citation checking, and answer organization collaborate in search-based AI applications.
Developer platform for experimenting with Gemini models, prompts, structured outputs, and API prototypes.
Good for understanding how model parameters, context, structured outputs, and evaluation connect when transitioning from prompt experimentation to API calls.
Microsoft's enterprise platform for building, evaluating, deploying, and managing AI applications and agents.
Good for explaining how enterprise-level models and agent projects require model directories, evaluation, deployment, identity, permissions, and observability capabilities.
AWS's managed models and generative AI platform, providing unified access to multiple foundation models and components needed for building Agents.
Suitable for explaining how cloud platforms provide a unified layer between model supply, access control, logging, deployment, and enterprise governance.
Google Cloud's machine learning and generative AI platform, providing model access, training, deployment, evaluation, and Agent tools.
Suitable for understanding model development, production deployment, permission governance, and evaluation within the same cloud platform context.
Anthropic's foundation model family. Models can be called via API or Claude application, with actual capabilities varying by model and product configuration.
Suitable for explaining the hierarchical relationship between model families, specific model versions, and user applications.
Anthropic's user-facing AI application, providing access to conversations, project contexts, files, and tools.
Suitable for comparing the differences between models and applications within the same organization, and how the product layer organizes contexts and tools.
An Agent environment for software engineering tasks, capable of reading codebases, executing commands, modifying files, and continuing work based on inspection results.
Suitable for explaining the relationships between Agents, tool calling, permissions, workflows, and verifiable results.
Google's foundation model family, callable via developer interfaces or Google products.
Suitable for explaining how the same model family can surface in different products such as search, productivity tools, developer tools, and third-party applications.
Google's user-facing AI application entry point, capable of connecting search, files, workspaces, and tool capabilities based on product and account.
Suitable for explaining why an application may or may not connect to external tools, which cannot be determined solely by the underlying model name.
SpaceXAI's foundation model family, accessible via developer API or Grok product entry.
Suitable for illustrating how the same model family can serve independent applications, social products, and other developer applications simultaneously.
SpaceXAI's user-facing Grok product entry point, providing conversation and other capabilities determined by the current product configuration.
Suitable for observing how the same model family is delivered to users through multiple surfaces such as web, mobile apps, and social platforms.
DeepSeek's foundational model family, with some models released as open weights, also available through official or other hosted endpoints.
Suitable for distinguishing between open weights, official applications, and third-party hosting, and explaining how the same model has different options across different runtime environments.
DeepSeek's user-facing AI application, providing conversation, files, and other product features. Specific capabilities are subject to current official product documentation.
Suitable for illustrating how institution names, model names, and application names often use the same terms, requiring confirmation of which layer is being referred to when reading.
A set of foundational model entry points provided by Zhipu, usable for conversation, reasoning, and other application scenarios.
Suitable for explaining the distinctions between model families, API services, and application products in the domestic model ecosystem.
The foundational model family provided by Alibaba Cloud, containing models of different scales and purposes, usable through cloud services, open weights, or application entry points.
Suitable for illustrating that a model family can have multiple scales and deployment methods, and cannot be fully characterized by a single product experience.
User-facing AI application, providing conversation, document processing, and related tool entry points.
Suitable for illustrating how application products organize user experience around context, files, and retrieval.
Agent environment for software engineering tasks, capable of performing multi-step modifications, testing, and result verification around codebases.
Suitable for observing how an Agent combines models, files, commands, permissions, and verification into a deliverable workflow.
Tencent's all-scenario AI office workbench. After the user presents a task, it can plan steps, invoke connectors and skills, generate files, and retain the process and results for review.
Suitable for illustrating that office Agents not only answer questions, but also combine task planning, tools, connectors, skills, and file delivery.
The officially open-sourced Agent Harness from DeepSeek, currently in developer preview. It treats models, tools, skills, conversations, sandboxes, storage, loops, scheduling, and interfaces all as replaceable plugins.
Suitable for clarifying the distinction between Harness and models: models are responsible for reasoning, while Harness is responsible for environment, tools, execution loops, lifecycle, and runtime trajectory.
A task-oriented general-purpose AI Agent product that can execute multi-step operations around a goal and deliver results.
Suitable for observing how general Agents organize goals, tools, execution steps, and final deliverables.
A development environment for code editing and project tasks, combining editor, model, codebase context, and command execution.
Suitable for illustrating how IDE products embed model capabilities into existing software engineering workflows.
GitHub's AI coding Agent for software development, combining codebase context, editor, commands, and code hosting workflows.
Suitable for illustrating how coding Agents are embedded in code hosting and review systems, rather than just appearing in standalone chat windows.
An open-source software engineering Agent that can read codebases, execute commands, modify files, and continue processing tasks based on test or error results.
Suitable for observing how open-source Agents advance software development tasks within inspectable code, command, and test loops.
A terminal-oriented AI pair programming tool that can edit files in the codebase, view diffs, and use Git to manage modifications.
Suitable for illustrating how coding Agents keep modifications inspectable through diffs, version control, and human confirmation.
Google's open-source terminal AI Agent, bringing Gemini capabilities to command-line tasks and development workflows.
Suitable for illustrating how Agent Harness combines models, terminal environments, tools, and task loops.
Google's agent-first development platform, including a desktop command center that can manage local Agents in parallel, a full IDE, CLI, and an SDK for building custom Agents.
Suitable for illustrating how the same Agent Harness can be exposed across desktop, IDE, command line, and SDK, and manage parallel tasks, workspaces, and scheduled tasks.
A development environment for repository tasks, providing capabilities such as conversation, editing, commands, and project context.
Suitable for comparing how different coding Agent products organize context, tools, and human approval.
An open-source coding Agent for code tasks, capable of working with editors, models, commands, and project files.
Suitable for illustrating how Agent products can be composed of editor extensions, model configurations, and tool permissions.
A tool for downloading and running open-weight models locally, providing local model management and invocation entry points.
Suitable for understanding boundaries such as local inference, model files, hardware resources, and data remaining on the local machine.
A local model inference project implemented in C/C++, supporting quantization and running various open-weight models.
Suitable for understanding the relationship between low-level inference, quantization, model files, and local hardware resources.
A local model running tool for desktop users, providing model download, parameter configuration, and local service entry points.
Suitable for observing how local model capabilities are influenced by model size, quantization, memory, and hardware.
A high-throughput inference engine for model service deployment, supporting inference optimizations such as batched requests and key-value caching.
Suitable for understanding the differences between inference services and regular chat applications, and how caching and service resources impact usage.