MicrosoftSpecialised AI models
Microsoft-Decision-1.1
Decision scoring · Classification · Routing
A specialised model for scoring predefined options in routing, classification and workflow control. Version 1.1 is generally available. Applications can use its scores to select a next step or request human review. Validate it on your workload; vendor benchmarks do not guarantee your results.
Access: Generally available; check current provider and deployment options
Official site / documentation for Microsoft-Decision-1.1 (opens in a new tab)OpenAISpecialised AI models
GPT-6 Cyber
Authorised security research · Vulnerability validation · Patch development
A specialised cybersecurity model for authorised vulnerability research, exploit validation and support for developing fixes. Use requires defined testing boundaries and human oversight.
Access: Initially API and Codex for selected approved Daybreak Red customers
Official site / documentation for GPT-6 Cyber (opens in a new tab)OpenAILanguage & reasoning
GPT
Text · vision · reasoning · coding
OpenAI’s general-purpose model family for reasoning, writing, analysis and coding. GPT-6.2 Astra and Sol support professional work, coding and computer use. Availability varies by product, plan and workspace settings. ChatGPT is an application using models; it is not the name of a single model.
Access: API; ChatGPT Work; Codex
Official site / documentation for GPT (opens in a new tab)AnthropicLanguage & reasoning
Claude
Text · vision · coding
Anthropic’s Claude portfolio includes Opus, Sonnet and Haiku, alongside Fable 5.2 and Mythos 5.2 for demanding technical work. Fable is broadly available; Mythos access requires approval through verification programmes. Choose using task evaluations, latency, cost and required controls; product and API access differ.
Access: Claude apps; API; cloud partners
Official site / documentation for Claude (opens in a new tab)Google DeepMindLanguage & reasoning
Gemini
Multimodal · reasoning · coding
Google’s multimodal family includes Gemini 4 Argon for coding, research and multi-step work, alongside Carbon for demanding reasoning, Hydrogen for fast agentic execution, and Neon for complex work with faster responses. The Gemini app, developer API and enterprise platform have different access and product terms.
Access: Gemini app; API; Google Cloud
Official site / documentation for Gemini (opens in a new tab)xAI / SpaceXAILanguage & reasoning
Grok
Reasoning · coding · multimodal
The model and assistant ecosystem developed by xAI, whose official site now uses SpaceXAI branding. Includes conversational, coding and media capabilities across different products.
Access: Grok app; API
Official site / documentation for Grok (opens in a new tab)Microsoft AILanguage & reasoning
MAI
Coding · cybersecurity · image · voice · transcription
Microsoft’s model portfolio includes MAI Code 2 and Code 2 Flash for coding, MAI Cyber 2 Flash for security, MAI Image 2.7 and Image 2.7 Flash for images, and MAI Realtime 1 for live speech. MAI names a portfolio; Copilot is a product that can use multiple providers. Access varies: Image 2.7 models are in public preview; Code 2 Flash has limited-access Foundry preview; Realtime 1 offers a Labs experience and registration of interest.
Access: Product-specific access; Foundry previews; MAI Labs
Official site / documentation for MAI (opens in a new tab)Microsoft ResearchLanguage & reasoning
Phi
Compact language models · reasoning
A Microsoft family of smaller language models designed for efficient deployment and reasoning. Phi is distinct from MAI; use each model card to check language support, license and intended use.
Access: Downloadable weights; hosting options
Official site / documentation for Phi (opens in a new tab)MetaLanguage & reasoning
Llama
Language · coding · multimodal variants
Meta’s model family used by developers to build and customize AI applications. Releases come with their own licenses and deployment requirements. Meta AI is a separate assistant experience.
Access: Downloadable weights; partner hosting
Official site / documentation for Llama (opens in a new tab)MetaLanguage & reasoning
Muse Wonder
Reasoning · coding · agent workflows
Meta’s Muse Wonder and Muse Wonder Full models support multi-step professional and coding tasks. They are distinct from the Llama family. Access differs by model; the announcement does not establish downloadable-weight availability.
Access: Muse Wonder: Meta Model API and Muse Code; Wonder Full: Muse Code subscribers
Official site / documentation for Muse Wonder (opens in a new tab)Google DeepMindLanguage & reasoning
Gemma
Open models · compact deployment
Google’s open model family for developer customization and deployment. Gemma and Gemini are different families with different access and licensing arrangements.
Access: Downloadable weights
Official site / documentation for Gemma (opens in a new tab)DeepSeekLanguage & reasoning
DeepSeek
Reasoning · coding · language
DeepSeek publishes language and reasoning models, including V4.1 Flash for visual understanding, coding and agent workflows, alongside V4 Pro. Its chat service and API can add search and document tools; these service capabilities are distinct from the model itself. Licences and deployment options vary by release.
Access: Chat; API; released model weights
Official site / documentation for DeepSeek (opens in a new tab)Alibaba / Qwen teamLanguage & reasoning
Qwen
Language · coding · vision · audio
Alibaba’s ecosystem includes language, coding, vision and audio models. Qwen4-Max adds native vision and long-running task execution for coding and professional work, alongside earlier Qwen families. Hosted model access does not imply downloadable weights; check the specific release and deployment terms.
Access: Chat; cloud APIs; released model weights
Official site / documentation for Qwen (opens in a new tab)Mistral AILanguage & reasoning
Mistral Models
Language · reasoning · coding · audio
Mistral’s portfolio includes Mistral Large 4 and Small 4 alongside Ministral, Magistral, Codestral, Devstral and Voxtral families for language, reasoning, coding and audio. Its catalogue distinguishes open-weight and commercial models, and also lists hosted third-party models that Mistral did not develop.
Access: API; selected downloadable weights
Official site / documentation for Mistral models (opens in a new tab)CohereLanguage & reasoning
Command, Embed & Rerank
Enterprise language · retrieval
Cohere develops Command language models alongside Embed and Rerank for enterprise retrieval. Embed 5 represents text and images for search; Rerank 5 orders retrieved results by relevance. Pro and Fast variants offer different quality and latency trade-offs. These components play different roles in RAG.
Access: API; private deployment options
Official site / documentation for Command, Embed & Rerank (opens in a new tab)IBMLanguage & reasoning
Granite
Enterprise language · task-specific models
IBM’s Granite portfolio includes Granite 4.2 language models alongside earlier releases and specialist speech, embedding and other task models. Language models support enterprise assistants, RAG and tool use. Consult the model selector for capabilities, hardware requirements and release-specific licensing.
Access: Open models; IBM ecosystem
Official site / documentation for Granite (opens in a new tab)Amazon / AWSLanguage & reasoning
Nova
Language · multimodal · speech · media
Amazon’s Nova portfolio includes Nova 2 Lite for everyday reasoning and automation, Nova 2.5 Sonic for real-time voice agents, and Nova Multimodal Embeddings for search across content types. The broader family spans multiple modalities. Nova models are distinct from Amazon Bedrock, which also hosts other developers’ models.
Access: AWS services
Official site / documentation for Nova (opens in a new tab)AI21Language & reasoning
Jamba
Long-context language
AI21’s Jamba research combines Transformer and state-space approaches for language modeling. AI21 also develops enterprise AI systems; check the current catalog for supported releases.
Access: Model releases; enterprise deployment
Official site / documentation for Jamba (opens in a new tab)NVIDIALanguage & reasoning
Nemotron
Reasoning · agents · specialist models
NVIDIA’s Nemotron portfolio includes Nemotron 3 Ultra for long-running reasoning tasks and 3.5 Super VL and Lightning for specialised multimodal and agent workloads. It provides models alongside training resources and deployment tools. Choose by task and infrastructure; check the licence and requirements of each release.
Access: Model weights; NVIDIA tools and hosting
Official site / documentation for Nemotron (opens in a new tab)Moonshot AILanguage & reasoning
Kimi
Reasoning · coding · knowledge work
Moonshot AI’s Kimi family includes K3 and K3.1 Pro. K3.1 Pro supports long-horizon coding, financial research, 3D creation and complex agent tasks. Kimi is also an assistant product; model, API and downloadable-release availability should be checked separately.
Access: Kimi app; API; research releases
Official site / documentation for Kimi (opens in a new tab)Z.ai / Zhipu AILanguage & reasoning
GLM
Language · coding · multimodal variants
Z.ai develops GLM language, coding and visual models. GLM-5.3-Flash adds native multimodal understanding for visual and agentic workflows within the GLM-5 family. Earlier GLM releases remain useful references; check each model’s current availability, deployment requirements and licence.
Access: Chat; API; released model weights
Official site / documentation for GLM (opens in a new tab)MiniMaxLanguage & reasoning
MiniMax
Language · agents · video · speech · music
MiniMax’s portfolio spans M3 and M3.1 Flash language models, H3 video and H3 Image, Speech 2.8 and Music 3.0. These families serve different text, visual and audio tasks. App, API and open-weight availability differ by model; one release does not replace every other capability.
Access: Apps; API; selected open-weight releases
Official site / documentation for MiniMax (opens in a new tab)TencentLanguage & reasoning
Hunyuan
Language · image · video · 3D
Tencent’s Hunyuan ecosystem spans language and visual generation. Its public repositories document individual model families rather than one universally available model.
Access: Official repositories; release-specific access
Official site / documentation for Hunyuan (opens in a new tab)BaiduLanguage & reasoning
ERNIE
Language · multimodal
Baidu’s ERNIE model ecosystem includes language and multimodal research and products. Official release notes are the starting point for checking current access and deployment options.
Access: Baidu services; model releases
Official site / documentation for ERNIE (opens in a new tab)Technology Innovation InstituteLanguage & reasoning
Falcon
Language · vision research
A model family developed by the UAE’s Technology Innovation Institute. Its official site documents language and perception releases, with access and licensing specific to each model.
Access: Published models and documentation
Official site / documentation for Falcon (opens in a new tab)Allen Institute for AI (Ai2)Language & reasoning
Olmo
Open language-model research
Ai2’s open language-model initiative emphasizes transparent research, including models and supporting development resources. Useful for studying how models are built as well as deploying them.
Access: Research models; training resources
Official site / documentation for Olmo (opens in a new tab)Liquid AILanguage & reasoning
Liquid Foundation Models
Compact models · on-device AI
Liquid AI develops compact models for local processing and constrained hardware. LFM3 adds text, vision, audio and combined-modality options alongside LFM2.5. Its d1.2 decision models address structured decisions. Evaluate the specific model, licence and hardware requirements for your deployment.
Access: Model releases; deployment tools
Official site / documentation for Liquid Foundation Models (opens in a new tab)Sarvam AILanguage & reasoning
Sarvam
Indian languages · speech · enterprise AI
Sarvam builds language, speech and document capabilities with a focus on Indian languages. Its portfolio includes Sarvam 30B and 105B, Saaras speech recognition, Bulbul speech generation and Sarvam Vision. Prism, built on Sarvam 105B, scores predefined answer options for structured decisions rather than generating free-form replies.
Access: Developer APIs; model releases; enterprise services
Official site / documentation for Sarvam (opens in a new tab)OpenAIImage & video
GPT Image
Image generation · editing
OpenAI’s image model family supports image generation and editing. Check the model catalog for supported inputs, outputs and access; image billing can differ from text-token billing.
Access: API; product-dependent access
Official site / documentation for GPT Image (opens in a new tab)Google DeepMindImage & video
Gemini Image
Image generation · editing
Google’s image generation and editing model offering, also associated with Nano Banana branding. It is a specialist capability within the broader Google AI ecosystem.
Access: Google products; developer services
Official site / documentation for Gemini Image (opens in a new tab)Google DeepMindImage & video
Veo
Video generation
Google DeepMind’s video generation model family. Product availability, video controls and supported outputs depend on the release and access channel.
Access: Google products; developer services
Official site / documentation for Veo (opens in a new tab)Stability AIImage & video
Stable Diffusion & Stability Models
Image generation · creative models
Stability AI develops generative models for creative workflows, including Stable Diffusion. Downloadable access does not mean every release has identical commercial-use terms.
Access: Hosted services; selected model weights
Official site / documentation for Stable Diffusion & Stability models (opens in a new tab)Black Forest LabsImage & video
FLUX
Image generation · editing
Black Forest Labs develops the FLUX image model ecosystem. Its releases offer different capabilities and licensing arrangements for creative and developer workflows.
Access: API; selected model weights
Official site / documentation for FLUX (opens in a new tab)MidjourneyImage & video
Midjourney
Visual generation
A visual AI research lab and creative service. Users primarily access its models through the Midjourney product rather than downloading a general-purpose language model.
Access: Hosted creative application
Official site / documentation for Midjourney (opens in a new tab)AdobeImage & video
Firefly
Creative generation · editing
Adobe’s generative AI model and product ecosystem supports creative workflows. The Firefly application can also offer partner models, so check which model a workflow uses.
Access: Adobe applications; developer offerings
Official site / documentation for Firefly (opens in a new tab)RunwayImage & video
Runway
Video · visual generation
Runway develops generative models and tools for visual storytelling and video workflows. Its platform combines model capabilities with production and editing features.
Access: Creative platform; developer offerings
Official site / documentation for Runway (opens in a new tab)ElevenLabsVoice & music
ElevenLabs
Voice generation · transcription · voice agents
An audio AI developer and platform for synthetic speech, transcription and conversational voice applications. Voice consent and usage rights remain important when using generated audio.
Access: Apps; APIs
Official site / documentation for ElevenLabs (opens in a new tab)OpenAIVoice & music
OpenAI Audio Models
Speech recognition · synthesis · realtime audio
OpenAI offers specialist transcription, speech-generation and realtime audio models. These have different inputs, outputs and endpoints; check the model documentation for supported capabilities and billing.
Access: API; product-dependent access
Official site / documentation for OpenAI audio models (opens in a new tab)CartesiaVoice & music
Sonic & Ink
Speech synthesis · speech recognition
Cartesia develops Sonic text-to-speech and Ink speech-to-text models, with infrastructure for interactive voice agents. These are specialist audio models rather than general chat assistants.
Access: API; deployment offerings
Official site / documentation for Sonic & Ink (opens in a new tab)DeepgramVoice & music
Deepgram
Speech recognition · synthesis · voice agents
Deepgram provides speech-to-text, text-to-speech and voice-agent infrastructure. It is useful for real-time audio applications and transcription workflows.
Access: APIs; enterprise offerings
Official site / documentation for Deepgram (opens in a new tab)SunoVoice & music
Suno
Music generation
A music generation platform that turns prompts into musical outputs. Check current subscription and usage terms before using generated music commercially.
Access: Hosted application
Official site / documentation for Suno (opens in a new tab)OpenAIAssistants & applications
ChatGPT & Codex
Knowledge work · coding · agents
ChatGPT is OpenAI’s user-facing AI workspace; Codex supports software development. These products combine models with tools and workflows, so an application’s capabilities are broader than its base model.
Access: Web and desktop products; plans vary
Official site / documentation for ChatGPT & Codex (opens in a new tab)AnthropicAssistants & applications
Claude & Claude Code
Knowledge work · software development
Claude is Anthropic’s assistant experience, while Claude Code supports development workflows. The products use Claude models and add tools, integrations and execution environments.
Access: Claude products; plans vary
Official site / documentation for Claude & Claude Code (opens in a new tab)MicrosoftAssistants & applications
Microsoft Copilot
Productivity · work context · agents
Microsoft’s Copilot product ecosystem brings AI into work and everyday tasks. Copilot can combine Microsoft and partner technology; it is not another name for the MAI model family.
Access: Microsoft products; licensing varies
Official site / documentation for Microsoft Copilot (opens in a new tab)GitHub / MicrosoftAssistants & applications
GitHub Copilot
Coding · development agents
An AI development assistant and agent product integrated with developer workflows. It provides access to models and tools, rather than representing one foundation model.
Access: GitHub products; plans vary
Official site / documentation for GitHub Copilot (opens in a new tab)PerplexityAssistants & applications
Perplexity & Sonar
AI search · grounded answers
Perplexity offers AI-assisted search and research. Its developer platform includes search and grounded-answer APIs; distinguish the search product from the underlying models and retrieval services.
Access: Search application; API platform
Official site / documentation for Perplexity & Sonar (opens in a new tab)WRITERAssistants & applications
WRITER
Enterprise agents · workflows
An enterprise AI platform for agentic work. Its value includes workflow and organizational integration; evaluate the platform’s controls and deployment options separately from model performance.
Access: Enterprise platform
Official site / documentation for WRITER (opens in a new tab)RekaAssistants & applications
Reka
Visual intelligence · enterprise workflows
An AI developer focused on visual intelligence and applications. Its official site describes the current product portfolio and access options.
Access: Enterprise products; developer offerings
Official site / documentation for Reka (opens in a new tab)MicrosoftEnterprise & inference platforms
Microsoft Foundry
Model catalog · agent development
A platform for building and operating AI applications with Microsoft and partner models. A model being listed in a catalog does not guarantee availability in every region or deployment type.
Access: Azure platform
Official site / documentation for Microsoft Foundry (opens in a new tab)AWSEnterprise & inference platforms
Amazon Bedrock
Multi-provider models · agents
A managed service for building generative AI applications using models from multiple providers. Bedrock is the platform; Nova is Amazon’s own model portfolio.
Access: AWS managed service
Official site / documentation for Amazon Bedrock (opens in a new tab)Google CloudEnterprise & inference platforms
Gemini Platform
Models · agent development · operations
Google Cloud’s platform for building and operating AI applications and agents. It combines models, development tools and operational capabilities; it is separate from the Gemini consumer app and employee-facing Gemini for business offering. Older documentation may use Vertex AI or Gemini Enterprise Agent Platform branding.
Access: Google Cloud platform
Official site / documentation for Gemini platform (opens in a new tab)IBMEnterprise & inference platforms
Watsonx
Enterprise AI · data · governance
IBM’s enterprise AI and data platform portfolio. Granite supplies models; watsonx provides capabilities for building, managing and governing AI solutions.
Access: IBM platform
Official site / documentation for watsonx (opens in a new tab)Together AIEnterprise & inference platforms
Together AI
Inference · training · model hosting
An AI infrastructure provider offering model inference and customization services. It hosts models from different developers; the hosting provider is not necessarily the model’s author.
Access: Cloud platform
Official site / documentation for Together AI (opens in a new tab)Fireworks AIEnterprise & inference platforms
Fireworks AI
Inference · customization
A platform for serving and specializing AI models. Useful when comparing deployment and customization options independently from the choice of foundation model.
Access: Cloud platform
Official site / documentation for Fireworks AI (opens in a new tab)GroqEnterprise & inference platforms
Groq
Fast model inference
An inference infrastructure provider serving supported AI models. Groq is different from Grok: Groq provides inference infrastructure, while Grok is the xAI model and product ecosystem.
Access: Inference platform
Official site / documentation for Groq (opens in a new tab)Luma AIImage & video
Ray & UNI
Video · image · creative workflows
Luma develops Ray video models and UNI image models. Its creative platform also integrates partner models; distinguish Luma’s own families from the other models offered in the app.
Access: Luma App; API
Official site / documentation for Ray & UNI (opens in a new tab)Kuaishou / Kling AIImage & video
Kling
Video · image · motion control
Kuaishou develops Kling AI visual generation models and a creative platform for images, video and motion control. The official site provides model guides, release history and developer access.
Access: Creative platform; API
Official site / documentation for Kling (opens in a new tab)ByteDance SeedImage & video
Seed Models
Multimodal research · creative generation
ByteDance’s Seed team develops models across several modalities, including Seedream image and Seedance video families. Its research site links releases and the relevant product or developer channels.
Access: Research releases; product-specific access
Official site / documentation for Seed models (opens in a new tab)No matching entries. Try a provider name or choose another category.