Reference Guide
Platform Capacity & UAP Analysis
A comprehensive reference guide to major AI platforms, core capabilities, and distinct architectural features across the generative ecosystem.
Notice: This guide is an informational capability and market awareness resource. It describes platform specializations and features. It is not an approved platforms list for PMG personnel or client engagements. For approved platforms and policy constraints, refer to Approved AI Platforms & Risk Assessment.
What is This Guide?
The AI platform landscape expands weekly. This guide provides teams with a curated, comparative breakdown of major model families, foundational capabilities, and technical differentiators.
Evaluation Fields
- Type: Modality / primary task focus.
- Provider & Region: Originating developer and jurisdiction.
- Specialties: Distinct competitive advantages.
- Key Points: Operational caveats & takeaways.
Select Edition:
2026 Q3
Published: 16 July 2026
Agentic & Autonomous Platforms
AI systems capable of autonomous execution, tool use, and multi-step goal completion without constant human prompting.
| Platform | Type | Provider | Region | Specialties & Capabilities | Key Points |
|---|---|---|---|---|---|
| Claude Computer Use & MCP | Agentic / OS Control | Anthropic | Autonomous desktop control, Model Context Protocol (MCP) v2.1 | Opus 4.7 enables native GUI interaction. MCP v2.1 standardizes tool use across platforms. | United States |
| Cursor Background Agents | Agentic / Code | Anysphere | Cloud-based autonomous coding, async multi-file refactoring | Agents run in the cloud, handling entire tickets or PRs while the developer does other work. | United States |
| Perplexity Computer | Agentic / Web | Perplexity AI | Autonomous web browsing, Custom Skills, form filling | Acts as a true web agent, executing multi-step workflows like booking flights or filling enterprise forms. | United States |
| Amazon Nova Act | Agentic / Automation | AWS | Browser automation, 90% reliability, low latency | Achieves 90% reliability in complex web navigation at 1/100th the cost of legacy RPA. | United States |
| Kimi Agent Swarm | Agentic / Multi-Agent | Moonshot AI | Multi-instance coordination, distributed task execution | K2.5+ architecture allows spawning multiple AI instances to collaborate on massive research or coding tasks. | Chinese |
Chat & General Assistants
General-purpose AI chat platforms. Good starting points for most tasks.
| Platform | Type | Provider | Region | Specialties & Capabilities | Key Points |
|---|---|---|---|---|---|
| Claude | Chat / Assistant | Anthropic | Long-document analysis, nuanced writing, complex reasoning | Claude Opus 4.7 is the current frontier. Features MCP v2.1 and Claude Design for Pro/Max/Team. | United States |
| ChatGPT | Chat / Assistant | OpenAI | Unified system with GPT-5.5 Instant, reduced hallucinations | GPT-5.5 and GPT-5.5 Pro released April 2026. GPT-5.5 Instant is the new default with improved personalization. | United States |
| Gemini | Chat / Multimodal | Google DeepMind | Native multimodal (text, audio, image, video) built on sparse MoE | Gemini 3.1 Pro dropped in Feb 2026. Deep integration across Google Workspace and Android. | United States |
| Grok | Chat / Search | xAI | Multi-agent collaboration system, real-time X data | Grok 4.20 Beta/Heavy released Feb 2026 with multi-agent sub-systems. Grok 5 (6T) expected later in 2026. | United States |
| Meta AI | Chat / Multimodal | Meta | Muse Spark integration across Meta apps | Meta Superintelligence Labs released Muse Spark in April 2026 as a complement to Llama. | United States |
| Character.AI | Chat / Roleplay | Character Technologies | Stories feature, PipSqueak model, DokiDoki Imagine | 20M MAU with 75 mins daily use. Introduced Charms currency and mid-chat image generation. | United States |
Enterprise & Developer Platforms
Platforms designed for business deployment, API access, and building AI-powered products.
| Platform | Type | Provider | Region | Specialties & Capabilities | Key Points |
|---|---|---|---|---|---|
| Microsoft Copilot | Enterprise / Chat | Microsoft | PowerPoint/Excel/Word Agents, Hey Copilot wake word | April 2026 brought 41 updates including custom MCP support and unified Teams experience. | United States |
| Cohere | Enterprise | Cohere | Multilingual & Sovereign AI, Rerank 4 | Hit $240M ARR and positioning for a 2026 IPO. Focus on multilingual models and enterprise search. | United States |
| Mistral | Chat / Code | Mistral AI | Mistral Large 3 (675B MoE), 256k context window | Mistral Large 3 is the largest open-weight MoE model. Mistral Small 4 (119B MoE) also released. | European |
Code & Development
AI tools built specifically for writing, reviewing, and working with code.
| Platform | Type | Provider | Region | Specialties & Capabilities | Key Points |
|---|---|---|---|---|---|
| Cursor | Code | Anysphere | Background Agents, Cloud Agents, Supermaven autocomplete | Evolved into a full AI development platform with MCP integrations and enhanced Hooks. Following the announcement of the Cursor acquisition by xAI, Cursor is immediately restricted and will be reconsidered as a possible vendor following an updated Data Processing Agreement and sufficient differentiation from xAI's high-risk posture. | United States |
| GitHub Copilot | Code | Microsoft | Autonomous PRs, Copilot Workspace, multi-model support | Now supports swapping between Claude, GPT, and Gemini. Autonomous coding agents handle entire issues. | United States |
| Qwen3-Coder | Code | Alibaba Cloud | Specialized for massive codebases, 128k+ context | Released alongside Qwen 3.7. Outperforms specialized coding models on SWE-bench. | Chinese |
Search & AI Browsers
AI platforms and browsers focused on finding, synthesising, and acting on real-time information.
| Platform | Type | Provider | Region | Specialties & Capabilities | Key Points |
|---|---|---|---|---|---|
| Perplexity | Search / Browser | Perplexity AI | Comet AI-native browser, Deep Research, Agentic Computer | 45M+ users. Launched Comet browser for iOS/Enterprise. Computer tools feature Custom Skills. | United States |
| ChatGPT Search | Search | OpenAI | Integrated into GPT-5.5, real-time news, publisher partnerships | Fully replaced legacy Bing chat interface. Deep links to premium publishers like WSJ and Forbes. | United States |
Image & Visual Creation
Platforms for generating, editing, and manipulating images with AI.
| Platform | Type | Provider | Region | Specialties & Capabilities | Key Points |
|---|---|---|---|---|---|
| Midjourney | Image / Creative | Midjourney Inc. | V7 standard, Omni Reference, Draft Mode, voice prompts | V7 is the current recommended version, replacing older parameters with Omni Reference. | United States |
| FLUX.2 | Image / Creative | Black Forest Labs | FLUX.2 (32B rectified flow transformer), 4MP output | FLUX.2 Pro delivers elite prompt adherence, hex color codes, and JSON prompting. | European |
| Adobe Firefly | Image / Creative | Adobe | Firefly AI Assistant, 4K cinematic video model | New creative agent enables workflow direction. 2026 Firefly Video Model creates 4K b-roll. | United States |
| Stable Diffusion 4 | Image / Creative | Stability AI | Open-source image generation, local deployment | SD 4.0 released with native FLUX-like architecture. Foundation for community fine-tunes. | European |
Video Creation
AI tools for generating and editing video content.
| Platform | Type | Provider | Region | Specialties & Capabilities | Key Points |
|---|---|---|---|---|---|
| Runway | Video / Creative | Runway AI | Gen-4 with World Consistency, up to 4K | Gen-4 released May 2026. Defining feature is World Consistency for 5-10s clips. | United States |
| Sora 2 | Video / Creative | OpenAI | Sora 2 flagship video/audio model | Standalone Sora app discontinued April 26, 2026. Sora 2 released as an integrated model in ChatGPT. | United States |
| Veo 3.1 | Video / Creative | Google DeepMind | Veo 3.1 high-resolution AI video | Veo 3.1 available via Vertex AI, competing directly with Runway Gen-4. | United States |
| Kling 2.0 | Video / Creative | Kuaishou | High-motion video, 2-minute generations | Dominating the Asian market. Unmatched physics and character consistency in long-form AI video. | Chinese |
Music & Audio Generation
AI tools for generating music, songs, and audio from text prompts.
| Platform | Type | Provider | Region | Specialties & Capabilities | Key Points |
|---|---|---|---|---|---|
| Suno | Music / Creative | Suno AI | v5.5 with AI Voice Recording, Custom Models, Studio DAW | v5.5 released March 2026. Features 44.1kHz audio, Style Reference, and up to 3 custom model variants. | United States |
| Udio | Music / Creative | Udio AI | Licensed era with major label deals | Shifted to a licensed model with deals with major labels and indie rights bodies for commercial safety. | United States |
| ElevenLabs | Audio / Voice | ElevenLabs | Voice cloning, sound effects, AI music | Expanded beyond voice into full sound effect generation and short music stems. Dominates TTS. | United States |
Open-Weight & Foundation Models
Models whose weights are publicly released, allowing local or private deployment.
| Platform | Type | Provider | Region | Specialties & Capabilities | Key Points |
|---|---|---|---|---|---|
| Llama 4 | Open Weights | Meta AI Research | Llama 4 Scout (10M context), 16 experts | Llama 4 Scout (109B total / 17B active) offers a 10M token context window. Behemoth still in training. | United States |
| Qwen 3.7 | Open Weights | Alibaba Cloud | Qwen 3.7 Agent Frontier, Qwen3-Coder-480B | Qwen 3.7 released May 2026. Flagship models include Qwen3-235B-A22B and specialized Coder variants. | Chinese |
| DeepSeek V4 | Open Weights | DeepSeek | DeepSeek V4 (1.6T MoE), 1M context, 10x lower cost | V4 Pro and V4 Flash available. Rivals GPT-5.5 and Claude Opus 4.7 at a fraction of the cost. | Chinese |
| Gemma 4 | Open Weights | Google DeepMind | Gemma 4 for edge and on-device AI | Gemma 4 released in April 2026, continuing Google's push for efficient on-device deployment. | United States |
Chinese & Asian Lab Models
Models and platforms developed by Asian AI labs; notable for their capabilities and rapid development pace.
| Platform | Type | Provider | Region | Specialties & Capabilities | Key Points |
|---|---|---|---|---|---|
| Kimi | Chat / Research | Moonshot AI | Kimi K2.5/K2.6/K2.7 Code, Agent Swarm | K2.5 (Jan 2026) introduced Agent Swarm. K2.6 (April) and K2.7 Code (June 2026) followed rapidly. | Chinese |
| MiniMax | Chat / Multimodal | MiniMax | MiniMax M2.5 model | M2.5 model released in China in March 2026, showcasing rapid regional advancements. | Chinese |
| Zhipu AI | Chat / Multimodal | Zhipu AI | GLM-5.1 model | GLM-5.1 released in April 2026, competing in the dense multimodal space. | Chinese |
| Kuaishou (Kling) | Video / Multimodal | Kuaishou | Kling 2.0 video and image generation | Rapid iteration in video AI. Kling 2.0 is heavily integrated into Chinese social and e-commerce apps. | Chinese |