Our offices

  • Exceev Consulting
    61 Rue de Lyon
    75012, Paris, France
  • Exceev Technology
    332 Bd Brahim Roudani
    20330, Casablanca, Morocco

Follow us

Preferences

Brand kit

Exceev Research · 2026 edition - The AI Tool Atlas

A field guide to the companies building the AI stack—from model labs and agent harnesses to coding tools, inference, evaluation, security, and creative engines.

Tools + product families
426
Market categories
15
Open-source entries
145

01 / Signal

Four truths hiding inside the noise.

01

The open layer is infrastructure

Frameworks, local runtimes, and agent harnesses are disproportionately open; consumer apps and hosted builders are overwhelmingly proprietary.

02

Open weights ≠ open source

A downloadable model can still restrict use, hide training data, and omit training code. The atlas keeps that label separate.

03

Harnesses are the new interface

The durable advantage is shifting from chat UI to the loop around the model: tools, permissions, memory, sandboxes, traces, and resumability.

04

Vibe coding is splitting in two

Prompt-to-app products optimize time-to-demo; coding agents optimize change quality inside real repositories. The overlap is growing, but the jobs differ.

02 / Directory

Search the stack.

Filter the complete directory by company, capability, category, or access model.

426 results

  • Open source
  • Open weights
  • Source available
  • Hybrid
  • Proprietary
Model labs & providers

OpenAI

OpenAI · Frontier model lab

ChatGPT · GPT APIs · GPT-6 Astra · Codex · Sora (web/app discontinued April 26, 2026; API September 24, 2026) · GPT-Live-1 · Agents SDK · Agents API · GPT-6 Sol · GPT-6 Luna · GPT-6.1 Sol · Dots · Decisions API (generally available; GPT-6 Luna) · GPT-Live-1.1 and GPT-Live-1.1 Quip (ChatGPT Voice) · GPT-6.1 Terra (OpenAI API) · GPT-6 Cyber (API and Codex; initially limited to selected approved Daybreak Red customers) · GPT-Image-2.6 Flare and Sunburst (API; image generation, editing and GIF creation)

HybridStatus: Active. Opens the official website in a new tab.
Model labs & providers

Anthropic

Anthropic · Frontier model lab

Claude · Claude Code · Claude Fable 5.1 · Claude Mythos 5.1 · Agent SDK · MCP · Model Hardware Standard · Claude Opus 5.5 · Claude Sonnet 5.5 · Claude Haiku 5.5 · Claude Duet (waitlisted research preview; US Pro and Max) · Claude Security (public beta; Enterprise) · Agent Skills (standard donated to the Agentic AI Foundation) · Messages API message threads (public beta; incremental conversation requests; token usage unchanged; not ZDR or HIPAA eligible) · Claude Fable 5.2 (Claude API and supported cloud platforms) · Claude Mythos 5.2 (limited to organizations approved through Anthropic verification programs)

HybridStatus: Active. Opens the official website in a new tab.
Model labs & providers

Thinking Machines Lab

Thinking Machines Lab · Multimodal model lab and model training platform

Inkling family (open weights) · Inkling 2 Small (downloadable weights since November 12, 2026; hosted through Tinker and other providers) · Tinker fine-tuning, sampling and Playground

HybridStatus: Active. Opens the official website in a new tab.
Model labs & providers

Google DeepMind

Google · Frontier model lab

Gemini · Gemini 3.8 Flash · Gemini 3.8 Flash Cyber · Veo · Imagen · Lyria · Gemma · Gemini 4 Argon (generally available) · Gemini 4 Argon Cyber Permissive (Fairwind allowlist) · Gemini 4 Carbon (paid API access; Google AI Ultra rollout in the Gemini app, Antigravity and AI Mode) · Genie 4 (research preview for trusted testers; coming soon to Google AI Ultra subscribers in Project Genie and early-access partners in the Gemini API and Google AI Studio) · Gemini Omni 1.2 Flash (GA in Gemini API and Gemini for business; native video generation and conversational editing up to 15 seconds) · Gemini 4 Hydrogen (Gemini API, Gemini for business and Gemini app) · Gemini Decisions API and Gemini Decision 1 (multimodal classification and scoring)

HybridStatus: Active. Opens the official website in a new tab.
Model labs & providers

Meta AI

Meta · Model lab

Llama · Muse Spark · Meta AI · Muse · SAM · AudioCraft · Muse Wonder (Meta Model API and Muse Code) · Muse Wonder Full (Muse Code subscribers) · Muse Spark Open (downloadable weights and hosted providers) · Muse Ember (multimodal model for tool calls, routing and coding subtasks; Meta Model API and Muse Code)

HybridStatus: Active. Opens the official website in a new tab.
Model labs & providers

Mistral AI

Mistral AI · Model lab + API

Vibe · Mistral models · Codestral · La Plateforme · Mistral Large 4 (Le Chonk, hosted API and downloadable weights under a custom license) · Voxtral 3 speech models (public preview) · Voice Agents SDK · Timestral forecasting models (public preview) · Mistral Studio (managed agents, workflows and apps)

HybridStatus: Active. Opens the official website in a new tab.
Model labs & providers

Inception

Inception (Inception Labs) · Diffusion language models and hosted APIs for reasoning and structured decisions

Mercury 3 (hosted API) · Mercury Voice (generally available for enterprise customers; enabled through sales) · Mercury Decide

ProprietaryStatus: Active. Opens the official website in a new tab.
Model labs & providers

SpaceXSI

SpaceX · Frontier model lab

Grok · Grok 4.7 · Grok API · Grok Bot · Grok Build · Grok Voice · Grok Imagine · Aurora · Grok Voice 3.0 (Voice Agent API; grok-voice-latest upgraded) · Text output watermarking (selected Grok API models from November 7, 2026; enabled by default; team admins can opt out per model) · Legacy Imagine image-quality and image-pro requests routed to grok-imagine-image-2.0-low (November 2, 2026)

HybridStatus: Active. Opens the official website in a new tab.
Model labs & providers

TypeSafe AI

TypeSafe AI · Structured decision model provider

Jev · System One Models · API

ProprietaryStatus: Active. Opens the official website in a new tab.
Model labs & providers

Liquid AI

Liquid AI · Device-native foundation models and multimodal decision models

Liquid Foundation Models · d1-3B (open weights) · d1-ultra (hosted decision API; text and image input) · LFM3 launch family (open weights): LFM3-1.7B, LFM3-VL-2B, LFM3-Audio-2B, LFM3-Omni-2B, d1-2B and d1-Omni-2B

HybridStatus: Active. Opens the official website in a new tab.
Model labs & providers

Reflection AI

Reflection AI · Open-weight models for coding, reasoning and agent workflows

Beam 1.0 (open weights) · Reflection API (public beta) · Mirror CLI coding harness

HybridStatus: Active. Opens the official website in a new tab.
Model labs & providers

Recursive

Recursive · Model lab releasing local coding models

Flint · local agentic coding · text reasoning and multi-turn tool use

Open weightsStatus: Active. Opens the official website in a new tab.
Model labs & providers

Glimpse

Elorian · Visual reasoning model for images and video

Glimpse (research preview) · API · Chat UI · Multimodal tool use

ProprietaryStatus: Active. Opens the official website in a new tab.
Model labs & providers

Clef

Cloudflare · Decision model family with typed probability outputs

Clef · Clef-flash · multimodal classification · Jev-compatible API · local weights · Workers AI inference · Clef-omni · audio and video decisions · Clef-flash hosted context: 24K tokens

Open weightsStatus: Active. Opens the official website in a new tab.
Model labs & providers

Clairic

Cloudflare · Typed extraction model with confidence estimates

Workers AI hosted preview · typed extraction · computations and estimates · forecasting (early checkpoint)

ProprietaryStatus: Active. Opens the official website in a new tab.
Model labs & providers

Cohere

Cohere · Enterprise model provider

Command · Embed · Rerank · North · North Small Translate

HybridStatus: Active. Opens the official website in a new tab.
Model labs & providers

DeepSeek

DeepSeek · Model lab + API

DeepSeek Chat · DeepSeek-V4.1-Flash · Reasoner · Coder · API · DeepSeek-V4.1-Flash-1017 (web, app and deepseek-flash API) · Search and Fetch APIs

HybridStatus: Active. Opens the official website in a new tab.
Model labs & providers

Qwen

Alibaba Cloud · Model family + assistant

Qwen · Qwen Chat · Qwen Code · Wan · Qwen3.8-2.4T-A95B (open weights) · Qwen3.8-Max (hosted API) · Qwen4-Max-Preview (Token Plan only; preview through November 25, 2026 UTC+8, then retires)

HybridStatus: Active. Opens the official website in a new tab.

03 / Method

What “all” means in a moving market.

In

Included

AI-native products, model providers, developer infrastructure, agent systems, and specialist generation platforms with an active public presence.

Out

Excluded

Conventional SaaS with a bolted-on AI feature, thin wrappers, abandoned projects, stealth companies, and every individual model checkpoint or SKU.

Read

License labels

Open source means code under an OSI-style license. Open weights and source available are separate. Hybrid combines a meaningful open component with a closed service.

Limit

Living index

No finite list can prove completeness in this market. This is a high-signal snapshot checked against official product pages and repositories, and maintained as a living index.

Research spine

Official product pages, documentation, repositories, changelogs, release feeds, and license files are the publication standard. Comparative indexes are used only to discover candidates.

A decision map, not a ranking. Classification describes access to the core product; a public repository alone does not make an entire product open source.