F.2 AI Background#

[1] Foundational Models#

Foundational models include private and open source services.

Private

1. GPT-5 (OpenAI-US): OpenAI’s flagship generation of reasoning and coding-optimized models, highly regarded for their mastery of software engineering benchmarks.

2. Claude (Anthropic-US): Renowned for their nuanced writing, sophisticated handling of complex logic, and deep integration with tool-use platforms like the Model Context Protocol.

3. Gemini 3 (Google-US) / Ultra: Google’s natively multimodal foundation models, designed for massive context windows, seamless deep integration across the Android/Google Workspace ecosystem, and heavy reasoning

4. Llama 4 (Meta-US): Meta’s flagship open-weight entry, heavily favored for building custom local AI workflows and broad general knowledge tasks.

5. Stable Diffusion (Stability AI-US): Open foundation models specialized in latent text-to-image and generative video editing.

6. Devstral (Mistral AI-EU): European-developed models celebrated for their efficiency, speed, and suitability for local development.

Open Source

7. DeepSeek (China): Open-weight with exceptional mathematical reasoning, efficient mixture-of-experts (MoE) architectures, and coding capabilities.

8. Qwen (Alibaba-China): A family of highly capable generalist and vision-language (VL) models developed by Alibaba, scoring highly across academic and reasoning benchmarks.

9. Mistral (EU and China): Created by former DeepMind and Meta scientists, Mistral offers highly capable, free-to-download models under the Apache 2.0 license.

10. Meta Llama (Deep Seek and Meta): highly customizable, and available in various sizes.

11. IBM Granite (US): IBM has open-sourced its enterprise-focused Granite family of language, code, and time-series foundation models under the permissive Apache 2.0 license.

12. Allen Institute for AI (AI2 - US): Based in Seattle, AI2 regularly open-sources its OLMo family of language models, providing full transparency into the training data and architecture.

13. Databricks DBRX (US): Released by the San Francisco-based data and AI company Databricks, DBRX is a powerful open-weights model utilizing a mixture-of-experts (MoE) architecture.


[2] Agents and UIs#

Private User Interfaces, Agents and Extensions

1. ChatGPT: The flagship interface for OpenAI GPT series (featuring GPT-4o and reasoning models), offering prompt chat, document analysis, voice interaction, and workspace canvases. Access it on OpenAI ChatGPT. ws. Access it on Anthropic Claude.

2. Gemini: Conversational AI interface, directly integrated with the Google Workspace ecosystem for exporting documents and generating real-time data. Access it on Google Gemini

3. Google AI Studio: A streamlined, web-based prototyping environment for developers to test prompts, adjust model parameters (temperature, token limits), and directly generate API code. Visit Google AI Studio.

4. Claude: Anthropic platform regarded for its advanced reasoning, large context windows, and specific interface features like “Projects” and artifact generation for coding and design.

5. Google Gemini: The premier UI for real-time information and deep integration with the Google Workspace ecosystem.

6. Perplexity AI: A specialized, citation-heavy UI designed specifically for research, deep web-diving, and source transparency

7. Cursor: It functions as a rebuilt VS Code fork designed for auto-completion, deep multi-file project context, and seamless “in-place” code edits.

8. Windsurf: Cognition desktop IDE. It stands out for its strong agentic workflow, autonomously planning, editing, and executing tests with reviewable diffs directly in your project.

9. GitHub Copilot: The most widely adopted tool. Best for teams seeking enterprise integration and real-time inline suggestions directly within their existing development tools.

10. Claude Code: Anthropic CLI-first agent. It excels at deep reasoning, reading your entire codebase, planning changes, and iterating on failures using natural language.Aider: A powerful command-line pair programmer. It is highly favored by terminal power users who want manual context control and transparent, incremental git commits.

11. Cline: A highly transparent, open-source VS Code extension and CLI. It allows developers to offload multi-file edits and terminal commands securely with in-editor diff approvals.

12. Codex: can perform tasks for you such as writing features, answering questions about your codebase, fixing bugs, and proposing pull requests for review.

Open Source User Interfaces, Agents and Extensions

13. Open WebUI Open WebUI is a self-hosted, extensible AI interface that supports Ollama, OpenAI API, and various LLM runners. It features granular permissions, responsive design, full Markdown/LaTeX support, hands-free voice/video calls, native Python function calling, persistent artifact storage, local RAG integration, web search, image generation/editing, and enterprise authentication. It’s ideal for privacy-focused deployments, local model experimentation, and enterprise use cases.

GitHub Link: open-webui/open-webui

14. LobeChat LobeChat is a modern, open-source ChatGPT-like UI with support for speech synthesis, multi-modal input, a plugin system, agent marketplace, and one-click deployment. It supports MCP (Model Context Protocol), smart internet search, chain of thought visualization, branching conversations, artifacts, file upload/knowledge base, multi-model providers, local LLMs (Ollama), visual recognition, TTS/STT, text-to-image generation, and custom themes. LobeChat excels in extensibility, user experience, and rapid deployment for both individuals and teams.

GitHub Link: lobehub/lobe-chat

15. Text Generation Web UI Text Generation Web UI is a popular open-source platform for running and interacting with local LLMs. It supports a wide range of models, offers a clean interface, and includes advanced features like model presets, plugins, and community-driven enhancements. It’s favored by developers and researchers for local experimentation and model benchmarking.

GitHub Link: oobabooga/text-generation-webui

16. Chatbot UI Chatbot UI is a Next.js-based, open-source chat interface that supports OpenAI and Azure APIs. It’s designed for easy self-hosting, customization, and rapid prototyping, making it a go-to choice for developers building custom chatbot experiences.

GitHub Link: mckaywrigley/chatbot-ui

17. LibreChat LibreChat is an open-source, multi-provider chat UI with support for MCP, rich UI features, and a focus on privacy. It’s designed for users who want a versatile, self-hosted chat experience with advanced agent orchestration and workflow management.

GitHub Link: danny-avila/LibreChat

18. AnythingLLM AnythingLLM is a multi-model, MCP-compatible UI that supports local and cloud deployment. It’s known for its flexibility, privacy, and ease of use, making it suitable for both personal and enterprise AI workflows.

GitHub Link: Mintplex-Labs/anything-llm

19. Flowise Flowise is a visual builder for LangChain-based LLM applications. It enables users to create complex agent workflows and pipelines with drag-and-drop ease, ideal for developers and non-developers alike.

GitHub Link: FlowiseAI/Flowise

20. LangFlow LangFlow is a visual pipeline builder for LangChain, allowing users to design, prototype, and deploy LLM-powered workflows. It’s perfect for rapid experimentation and building complex agent logic.

GitHub Link: logspace-ai/langflow

21. assistant-ui assistant-ui is a React/TypeScript-based component library for building custom LLM chat interfaces. It’s highly modular and customizable, making it a favorite for developers building tailored agent experiences.

GitHub Link: assistant-ui/assistant-ui

22. Streamlit Streamlit is a Python-based framework for quickly building and deploying interactive web apps, including LLM chat interfaces. It’s ideal for rapid prototyping, data science, and AI experimentation.

GitHub Link: streamlit/streamlit

23. Hugging Chat: An open-source, community-driven interface that hosts multiple foundational models (such as Mistral and Llama) in a single chat environment. Try it on Hugging Chat.

Github Link: huggingface/chat-ui