A structured reference map of the AI vendor landscape, open model ecosystem, and best tools by work type. Published every Saturday
| Capability Layer | OpenAI | Anthropic | Meta | Microsoft | Popularity | |
|---|---|---|---|---|---|---|
| Core Models | Gemini 2.5 Pro, Gemini 2.0 Flash | GPT-4.1, o3, o4-mini | Claude 3.7 Sonnet, Claude 3.5 Haiku | Llama 4 Scout, Llama 4 Maverick | Phi-4, Phi-4-mini | |
| APIs / Platform | Vertex AI, Google AI Studio | OpenAI API, Azure OpenAI | Anthropic API, Claude.ai Teams | Llama API, Meta AI Platform | Azure AI Foundry, Azure OpenAI Service | |
| Dev Tools | Vertex AI SDK, Gemini CLI | OpenAI SDK, Assistants API | Anthropic SDK, Workbench | Llama Stack, PyTorch | Semantic Kernel, Prompt Flow | |
| Assistants / Apps | Gemini App, NotebookLM | ChatGPT, ChatGPT Operator | Claude.ai, Claude Projects | Meta AI, WhatsApp AI | Microsoft Copilot, Copilot Studio | |
| Research Tools | NotebookLM Plus, Gemini Deep Research | ChatGPT Deep Research, o3 | Claude Research Mode, Claude 3.7 Sonnet | Llama 4, Meta FAIR Research | Microsoft Research AI, Copilot for Research | |
| Creative / Media | Imagen 4, Veo 3 | DALL-E 3, Sora | Claude Creative Writing | Emu Video, Meta Imagine | Designer, Copilot Image Creator | |
| Edge / On-device | Gemini Nano, MediaPipe | GPT-4o mini | Claude 3.5 Haiku | Llama 4 Scout, Llama 3.2 1B | Phi-4-mini, ONNX Runtime | |
| Voice / Speech | Gemini Live, Google Cloud TTS | Voice Mode, Whisper, TTS API | Claude Voice (beta) | SeamlessM4T, Voicebox | Azure Speech Services, Copilot Voice | |
| Agent Frameworks | Agent Development Kit, Vertex AI Agents | Responses API, Swarm | Claude Tool Use, Model Context Protocol | Llama Stack Agents, AnyTool | AutoGen, Semantic Kernel Agents |
| Vendor | Models | License | Practical Use |
|---|---|---|---|
| Meta | Llama 4 Scout 17B, Llama 4 Maverick 17B, Llama 3.3 70B | Llama Community License (open weights) | General chat, coding, RAG, local deployment, fine-tuning |
| Gemma 3 27B, Gemma 3 12B, Gemma 3 4B | Gemma Terms of Use (open weights) | Edge inference, fine-tuning, on-device assistants, multimodal tasks | |
| Mistral | Mistral Small 3.1, Mixtral 8x22B, Mistral 7B | Apache 2.0 / Mistral Research License | European compliance, RAG, enterprise fine-tuning, multilingual tasks |
| DeepSeek | DeepSeek-V3, DeepSeek-R1, DeepSeek-Coder-V2 | MIT License | Coding, reasoning, math, cost-effective API alternative |
| Alibaba / Qwen | Qwen2.5 72B, Qwen2.5-Coder 32B, QwQ-32B | Apache 2.0 / Qwen License | Multilingual (Chinese/English), coding, reasoning, long-context tasks |
| Microsoft | Phi-4 14B, Phi-4-mini 3.8B, Phi-3.5-MoE | MIT License | Edge deployment, constrained compute, STEM reasoning, fine-tuning |
| 01.AI / Yi | Yi-1.5 34B, Yi-1.5 9B | Apache 2.0 | Long-context tasks, multilingual, research baselines |
| Allen AI | OLMo 2 7B, OLMo 2 13B | Apache 2.0 (fully open including data) | Research, reproducibility, academic benchmarking |
| xAI | Grok-3 (weights TBD), Grok-2 12B | xAI Community License | Real-time web data, reasoning, chat |
Includes models actively used or relevant as of Saturday, 2 May 2026. Experimental or pre-release models excluded.
| AI Work Type | Best Closed (Speed / Quality) | Best Open (Control / Cost) | When to Use |
|---|---|---|---|
| Chat | ChatGPT (GPT-4.1), Claude 3.7 Sonnet | Llama 4 Maverick, Qwen2.5 72B | Use closed for highest quality and reliability; use open for privacy, self-hosting, or cost control |
| Coding | GPT-4.1, Claude 3.7 Sonnet | DeepSeek-Coder-V2, Qwen2.5-Coder 32B | Use closed for complex multi-file projects; use open for air-gapped or proprietary codebases |
| RAG | Claude 3.7 Sonnet, Gemini 2.5 Pro | Llama 3.3 70B, Mistral Small 3.1 | Use closed for long-context accuracy; use open with LlamaIndex or LangChain for on-prem document stores |
| Agents | ChatGPT Operator, Claude with MCP | Llama 4 via Llama Stack, DeepSeek-V3 via AutoGen | Use closed for production reliability and tool ecosystems; use open for custom pipelines and cost at scale |
| Search | Perplexity Pro, ChatGPT Search, Gemini Deep Research | SearXNG + Llama 4, Ollama + DuckDuckGo | Use closed for real-time web research; use open for private intranet search or GDPR-sensitive environments |
| Image | DALL-E 3, Imagen 4, Midjourney v7 | Stable Diffusion 3.5, FLUX.1 | Use closed for ease of use and quality; use open for unlimited generation, fine-tuning, and commercial control |
| Video | Sora, Veo 3, Runway Gen-4 | CogVideoX, Open-Sora | Use closed for high-quality short-form video; use open for research or when content licensing is a concern |
| Speech | OpenAI Whisper API, Azure Speech, Google Cloud TTS | Whisper large-v3, Coqui XTTS, Kokoro TTS | Use closed for enterprise SLA and streaming; use open for offline transcription, voice cloning, and low latency edge |
| Edge | Gemini Nano, GPT-4o mini API | Phi-4-mini, Llama 3.2 3B, Gemma 3 4B | Use open models for fully offline mobile and IoT deployment; use closed nano models when cloud fallback is acceptable |
| Fine-tuning | OpenAI Fine-tuning API, Vertex AI Tuning | Llama 4 Scout, Mistral 7B, Phi-4-mini via Axolotl or Unsloth | Use closed APIs for quick domain adaptation without infra; use open for full data ownership and repeated iteration |
| Multimodal | Gemini 2.5 Pro, GPT-4.1, Claude 3.7 Sonnet | Llama 4 Scout, Gemma 3 27B, Qwen2.5-VL 72B | Use closed for best vision-language performance; use open for document analysis pipelines requiring data privacy |
Closed = prioritises performance and ease of use. Open = prioritises control and cost. Updated weekly.