Comprehensive List of AI Tools

Welcome to a curated collection of tools, models, and platforms across the rapidly evolving landscape of artificial intelligence (AI) and machine learning (ML). This repository has been actively maintained since 2023 and aims to provide a reliable, up-to-date reference for the AI ecosystem.

This repository focuses on the most relevant systems across:

• large language models (LLMs)
• multimodal generation (image, video, audio)
• AI agents and automation
• developer platforms and APIs
• infrastructure and supporting tools

It is designed as a practical resource for developers, professionals, and enthusiasts looking to explore tools, understand the ecosystem, and build real-world AI applications.

A Living, Community-Driven Guide

This is a dynamic resource, continuously updated to reflect the fast-moving AI landscape. It is curated through community contributions, ongoing research, and iterative cleanup to maintain a high-signal and relevant collection.

Please Note:

• Independence: This is an independent, community-driven project. We are not affiliated with, sponsored by, or endorsing any of the tools listed.
• Accuracy: The AI space evolves rapidly. Links, pricing, and features may change — always verify details on official sources.
• Due diligence: Tools vary in quality, safety, and maturity. Please conduct your own research before using them.
• No guarantees: Some tools may be outdated, low quality, or unreliable — always double-check.

Ready to contribute? Before contributing, please review the Contribution Guidelines.
This helps keep the list consistent and high quality.

Note: The repository is currently undergoing a quality review and cleanup phase.
Submissions are still welcome, but low-effort or duplicate pull requests may be declined.

Found a broken link or something missing? Feel free to open an issue or submit a pull request.

🙌 Contributors

🔗 Navigation Table

Category	Description
🧠 Major AI Platforms & Model Ecosystems	Overview of leading AI platforms and model ecosystems including OpenAI, Google, Anthropic, Meta, Mistral, and xAI.
🤖 OpenAI	GPT models, ChatGPT, Codex, and OpenAI’s API ecosystem.
🔍 Google	Gemini models, Veo, Imagen, and Google’s AI ecosystem.
🤖 xAI / Grok	Grok models, real-time data integration, and multi-agent systems.
🧠 Anthropic	Claude models focused on long-context reasoning and reliability.
💻 Claude Code	Agentic coding system built for full software engineering workflows.
🧑‍💻 OpenAI Codex	Autonomous coding agent for development, debugging, and workflows.
🧑‍💻 Perplexity	Web-grounded AI system for search, multi-step reasoning, and source-based answers.
🇨🇳 Leading Chinese AI Models	Powerful and popular large language models from China.
🪦 AI Graveyard	Discontinued, retired, or replaced AI tools and models
🦞 OpenClaw Ecosystem	OpenClaw, ZeroClaw and related autonomous AI agent frameworks.
🤖 Other AI Agents	Collection of notable AI agent frameworks and autonomous systems beyond the major ecosystems.
🛠️ AI Agent Tools	Supporting tools, libraries, and research resources for building and running AI agents.
🎓 Free Stuff for Students & Educators	Free access to professional-grade tools, software, and cloud credits.
🤖 AI-Driven Robots and Robotic Dogs	Cutting-edge robotics powered by AI.
🤖 AI-Related Tools	Miscellaneous tools leveraging AI technology.
🆕 Additional AI and Productivity Tools	New AI tools and services for productivity and innovation.
💾 GPU Rental Services	Services for renting GPUs to train and deploy AI models.
🌐 Chrome Extensions	Useful AI-powered Chrome extensions.
🖋️ Learning AI/ML	Resources for learning AI and machine learning.
Image Generation Tools	Tools for generating high-quality AI-driven images.
Custom GPT Sites	Custom GPT-powered websites and platforms.
Video Generation Tools	AI tools for generating and editing videos.
Sound Generation Tools	AI tools for creating and modifying sound.
♫ Music Generation Tools	Tools for composing music using AI.
Voice-Related AI Tools	Applications for voice synthesis and enhancement.
Media Enhancement Tools	AI tools for enhancing media quality.
AI for Photo Editing	AI tools designed for professional photo editing.
🖼️ Image Prompting Tools	Tools to create better prompts for image generation models.
API Integration Sites	Websites offering API integrations for AI models.
🛠️ AI Compilers/Code Assistants	Tools for coding and compiling with AI assistance.
🛠️ LLM Ops	Operations and infrastructure for large language models.
🛠️ Consumer AI Hardware	Hardware designed for running AI applications at home.
💻 Software for Large Language Models	Software for training and using large AI language models.

Your contributions and suggestions are welcome to keep this list comprehensive and up-to-date!

---

---

🧠 Major AI Platforms & Model Ecosystems

🔝 Back to Top

This section covers the leading AI platforms and model ecosystems used globally, including both proprietary and open-weight systems. These platforms provide state-of-the-art capabilities across reasoning, coding, multimodal generation, and agent-based workflows.

---

OpenAI

A leading AI company focused on advanced multimodal and agent-based systems. OpenAI develops ChatGPT and a widely used API platform for text, image, audio, and computer-use models.

• ChatGPT - The primary interface for interacting with OpenAI models, supporting chat, coding, data analysis, research, and agent workflows.

• Flagship Models (Current):

  • GPT-5.5 – Latest flagship model for advanced reasoning, coding, research, data analysis, and agentic workflows.
  • GPT-5.5 Thinking – Reasoning-focused version available in ChatGPT for harder multi-step tasks.
  • GPT-5.5 Pro – Higher-accuracy variant for demanding professional and research workloads.
  • GPT-5.4 – Previous flagship model, still relevant as part of the GPT-5.x model family.
  • GPT-5.4 Pro – Previous high-performance variant for complex reasoning and professional work.
  • GPT-5.4 Mini / Nano – Smaller variants optimized for speed, cost-efficiency, and scale.

• Model Direction:

  • GPT-5.5 builds on GPT-5.4 with stronger coding, reasoning, tool use, and agentic task performance.
  • OpenAI continues consolidating reasoning, coding, multimodal, and agent capabilities into the GPT-5.x family.
  • Earlier variants are gradually being replaced, absorbed, or positioned as smaller/faster options.

• Multimodal & Specialized Models:

  • gpt-image-1 – Current image generation model with strong editing and prompt adherence.
  • gpt-realtime / gpt-audio – Models for speech, transcription, and real-time interaction.
  • Computer use models – Models designed for operating software interfaces and completing agentic computer tasks.

• API Platform:

  • Supports Responses API and Chat Completions API
  • Key capabilities:
    • Tool / function calling
    • Structured outputs
    • Agent workflows
    • Computer use and automation
    • Long-context support, including very large context windows depending on model

• Pricing Tiers:

  • Free – Limited ChatGPT usage
  • Plus ($20/month) – Higher limits and access to more capable models
  • Pro ($100/200 month) – Higher usage limits, access to the most capable models, and more advanced tools such as Deep Research and Codex
  • Business / Enterprise / Edu – Team and organization plans with higher limits and admin controls
  • API – Pay-as-you-go access across available model families

OpenAI Codex

OpenAI Codex is an agent-based coding system built on top of the latest GPT-5 model family. It has evolved from a code-generation model into a full software engineering platform capable of writing, editing, running, and reviewing code autonomously.

• Codex - OpenAI’s platform for agentic coding workflows and automated software development tasks.

• Core Models:

  • GPT-5.3-Codex – Advanced Codex-native agentic coding model for software engineering, terminal use, and broader computer-based workflows
  • GPT-5.2-Codex – Coding-focused GPT-5.2 variant optimized for long-horizon development, refactoring, migrations, Windows workflows, and cybersecurity-related tasks
  • GPT-5-Codex – Earlier GPT-5-based Codex model optimized for agentic coding tasks in Codex and similar environments
  • Codex variants (fast / low-latency) – Optimized versions for interactive coding and rapid iteration

• Capabilities:

  • Full codebase understanding – Reads, edits, and navigates large repositories
  • Autonomous development tasks – Can implement features, refactor code, and manage workflows
  • Testing & debugging – Runs tests, identifies issues, and proposes fixes
  • Tool & environment interaction – Works with terminals, files, and developer tools
  • Parallel execution – Handles multiple coding tasks simultaneously
  • Code review – Reviews changes and can identify bugs before deployment

• Agent Platform:

  • Runs tasks in sandboxed environments with access to project files
  • Supports long-running and multi-step workflows
  • Available via:
    • ChatGPT
    • Codex CLI
    • IDE integrations
    • Codex web
    • GitHub integrations
    • Developer APIs / SDKs where supported

• Positioning:

  • A shift from “code assistant” → autonomous software engineering agent
  • Competes with tools like Claude Code and GitHub Copilot Agents
  • Increasingly integrated into the broader OpenAI ecosystem through ChatGPT, Codex, and the API platform

---

Google

Google is a powerhouse in AI research and infrastructure, building the Gemini family and integrating AI deeply across Search, Workspace, Android, Cloud, and creative tools.

• Google Gemini - Google’s main conversational interface for interacting with its latest AI models.

• Flagship Models:

  • Gemini 3.5 Flash – Google’s current stable Gemini 3.5 model for agentic tasks, coding, reasoning, and fast multimodal workflows.
  • Gemini 3.1 Pro – Higher-end Pro model for complex workflows, stronger reasoning, and demanding real-world tasks.
  • Gemini 3 Flash – Fast frontier model optimized for speed, cost-efficiency, and everyday use.
  • Gemini 3.1 Flash-Lite – Lightweight model optimized for high-volume and cost-sensitive use.
  • Gemini 3.1 Flash Live – Built for more natural real-time audio and live multimodal interaction.

• Creative Models:

  • Veo 3.1 – Google’s current video generation model family, with stronger control, better consistency, and native audio support.
  • Imagen 4 / Nano Banana Pro / Nano Banana 2 – Google’s current image generation and editing models, focused on quality, speed, prompt adherence, and higher-end visual creation.
  • Lyria 3 / Lyria 3 Pro / Lyria 3 Clip – Google’s latest generative music models for creating custom tracks, loops, previews, and soundtracks.

• Open / Local Models:

  • Gemma 4 – Google’s newest open model family, built for advanced reasoning and agentic workflows, and designed to run well on consumer GPUs.
  • Gemma 4 12B – Mid-sized multimodal Gemma model designed for local and laptop-friendly agentic workflows.
  • Gemma 3 / 3n – Earlier lightweight open models still relevant for efficient and local-first deployments.
  • FunctionGemma / ShieldGemma / EmbeddingGemma – Specialized open models for tool use, safety classification, and embeddings.

• Key Features:

  • Native multimodality – Strong support across text, image, audio, and video.
  • Massive context – Built for long-context reasoning and large-scale document / code analysis.
  • Deep ecosystem integration – Embedded across Google Search, Workspace, Android, Cloud, and creative products.
  • Creative tooling – Google now has serious first-party image, video, and music generation instead of only chat-focused models.

• Developer Platform:

  • Vertex AI - Google Cloud’s enterprise AI platform for accessing Gemini, Imagen, Veo, Gemma, embeddings, and broader MLOps tooling.
  • Google AI Studio - Lightweight environment for quickly testing Gemini and other Google models.

---

xAI Grok

🔝 Back to Top

xAI develops the Grok family of models and multimodal systems, with a focus on real-time information, reasoning, and large-scale deployment. The ecosystem is tightly integrated with the X platform and accessible via API.

• xAI — Company → News · API docs · API console · GitHub
• grok.com — Main chat interface (web, iOS, Android), also integrated into X

---

Models

• Grok 4.3 — Current flagship API model

  • Model name:
    • grok-4.3
  • Supports text + image input, configurable reasoning, function calling, structured outputs, and agentic tool use
  • Up to 1M token context
  • Recommended default model for most API workloads

• Grok 4.20 API variants — Current specialized API model variants

  • Includes:
    • grok-4.20-0309-reasoning
    • grok-4.20-0309-non-reasoning
    • grok-4.20-multi-agent-0309
  • Supports reasoning, non-reasoning, and multi-agent modes
  • Up to 1M token context
  • Designed for coding, math, agents, and complex workflows

• Grok Build 0.1 — Coding model / agent system

  • Model name:
    • grok-build-0.1
  • Built for agentic coding workflows
  • Powers Grok Build CLI and developer workflows
  • Up to 256k token context

• Grok 4 Heavy — Advanced consumer variant

  • Uses multi-agent parallel reasoning for harder problems
  • Highest capability tier in Grok subscription plans

• Grok 4.1 Fast variants — Previous fast API generation

  • grok-4-1-fast-reasoning
  • grok-4-1-fast-non-reasoning
  • Previously optimized for low latency and cost-sensitive workloads
  • Retired API slugs now redirect to Grok 4.3

• Grok 3 — Previous generation

  • Introduced DeepSearch, Think mode, and large-scale training
  • Retired API slug now redirects to Grok 4.3

• Grok 1 — Open-weight release

  • ~314B MoE base model
  • Apache 2.0 → Hugging Face

---

Multimodal

• Grok Imagine — Image + video generation system

  • Models:
    • grok-imagine-image
    • grok-imagine-image-quality
    • grok-imagine-video
    • grok-imagine-video-1.5-preview
  • Supports generation + editing workflows
  • Available via API and Grok interface

• Aurora — Earlier image-generation model

  • Autoregressive image model previously used for Grok image generation
  • Focused on photorealism and multimodal input
  • Superseded by Grok Imagine for current image / video API workflows

---

Agent & Platform Capabilities

• Real-time data integration — Native access to live web + X data through search tools
• Agent workflows — Tool use, browsing, code execution, and multi-step execution
• Multi-agent systems — Exposed through Grok Heavy and multi-agent model variants
• Grok Build CLI — Terminal-based coding agent for software engineering workflows
• Tooling:
  • Web search
  • X search
  • Code execution
  • Retrieval / document processing
  • Remote MCP tools

---

API Pricing (USD, official)

• Grok 4.3 / Grok 4.20 variants:

  • $1.25 / 1M input tokens
  • $0.20 / 1M cached input tokens
  • $2.50 / 1M output tokens

• Grok Build 0.1:

  • $1.00 / 1M input tokens
  • $0.20 / 1M cached input tokens
  • $2.00 / 1M output tokens

• Grok Imagine:

  • Image generation from $0.02 / image
  • Higher-quality image generation up to $0.07 / image
  • Video generation from $0.05 / second depending on model and resolution

• Voice / Audio:

  • Voice Agent API — $0.05 / min (~$3/hour)
  • Text-to-Speech — $15.00 / 1M characters
  • Speech-to-Text — from $0.10 / hour

---

Subscriptions & Plans

• Free tier

  • Limited queries and features

• SuperGrok Lite

  • Lower-cost standalone Grok plan

• SuperGrok (~$30/month)

  • Full Grok access
  • Higher limits and priority usage

• SuperGrok Heavy

  • Access to Grok Heavy (multi-agent)
  • Maximum limits and advanced features

• X Premium (~$8/month)

  • Enhanced Grok access bundled with X features

• X Premium+ (~$40/month)

  • Higher Grok access bundled with X features

• Grok Business

  • Team-based access and admin controls

• Enterprise

  • Custom pricing / deployment

---

Anthropic

🔝 Back to Top

Anthropic focuses on building reliable, interpretable, and safety-aligned AI systems. It develops the Claude family of models, widely used for long-context reasoning, coding, and enterprise workflows.

• Claude.ai - The main interface for interacting with Claude models, supporting chat, documents, coding, and agent-style workflows.

• Current Models (Primary):

  • Claude Opus 4.8 – Anthropic’s most capable public model, designed for complex reasoning, advanced coding, agent workflows, and long-running task execution.
  • Claude Sonnet 4.6 – Default model balancing performance, speed, intelligence, and cost.
  • Claude Haiku 4.5 – Fastest model optimized for quick responses and high-throughput tasks.

• Still Available / Selectable Models:

  • Claude Opus 4.7 – Previous flagship, still available for compatibility and specific workloads.
  • Claude Opus 4.6 – Earlier Opus model for deep reasoning, coding, and long-running agent workflows.
  • Claude Opus 4.5 – Previous flagship, still available for compatibility and specific workloads.
  • Claude Sonnet 4.5 – Earlier balanced model, widely used in production systems.

• Capabilities:

  • Up to ~1M token context window – Enables processing of very large documents, codebases, and multi-step workflows.
  • Extended / adaptive thinking – Optional deeper reasoning and effort control for complex tasks.
  • Agent workflows – Designed for multi-step execution, tool use, and long-running processes.

• Model Structure:

  • Opus – Maximum capability and reasoning depth.
  • Sonnet – Balanced performance and efficiency.
  • Haiku – Speed and low latency.

• Next-Generation / Restricted:

  • Claude Mythos Preview – Invitation-only research preview model for defensive cybersecurity workflows.
  • Part of Anthropic’s Project Glasswing.
  • No public self-serve access.

• Key Differentiators:

  • Constitutional AI – Safety-first training approach.
  • Industry-leading long context – Strong performance on large inputs.
  • Enterprise reliability – Consistent, structured outputs for real-world workflows.

Claude Code

Claude Code is Anthropic’s agentic coding system designed to handle full software engineering workflows. Unlike traditional code assistants, it operates directly inside your development environment and can execute multi-step tasks autonomously.

• Claude Code – AI coding agent that works in your terminal, IDE, desktop app, or browser.

• Core Models:

  • Uses Claude model-family aliases such as Opus, Sonnet, and Haiku, resolving to the latest available models depending on configuration and task needs.

• Capabilities:

  • Full codebase understanding – Reads and maps entire repositories automatically.
  • Multi-file editing – Makes coordinated changes across large codebases.
  • Autonomous workflows – Can implement features, fix bugs, and handle tasks end-to-end.
  • Command execution – Runs terminal commands, scripts, and developer tools.
  • Tool integration – Works with Git, CI/CD, databases, and external services.
  • Parallel agents – Supports multi-agent workflows and background agents.

• Interfaces & Usage:

  • Terminal (primary interface)
  • VS Code / JetBrains integrations
  • Desktop app
  • Web interface
  • Slack / team workflows

• Agent System:

  • Uses tool-based execution (files, shell, APIs, web).
  • Supports long-running tasks and multi-step reasoning.
  • Can operate with memory/context across workflows.

• Platform Features:

  • Code Review / CI workflows – Automates pull request review and issue triage through GitHub Actions or GitLab CI/CD.
  • GitHub / GitLab integration – Reads issues, writes code, and opens PRs.
  • Configurable agent behavior – Supports custom instructions, skills, hooks, and workflows.

• Positioning:

  • A shift from “code assistant” → autonomous software engineering agent.
  • Competes directly with:
    • OpenAI Codex
    • Cursor (agent mode)
    • GitHub Copilot Agents
  • Particularly strong in long-context coding and large codebase understanding.

---

Meta AI

🔝 Back to Top

Meta is a leading force in open-weight AI, focused on making powerful foundation models widely accessible. Its Llama family is one of the most adopted model ecosystems for local, research, and commercial use.

• Meta Llama - Official hub for Llama models, research, and releases.

• Flagship Models (Current):

  • Llama 4 Scout – Current open-weight multimodal model optimized for efficiency, long context, and broad deployment
  • Llama 4 Maverick – Current higher-capability open-weight multimodal model for reasoning, coding, and assistant-style use
  • Llama 4 Behemoth – Previewed large teacher model used to improve the Llama 4 family
  • Llama 3.3 70B – Strong previous-generation text model offering performance close to larger Llama 3.1 models at lower serving cost

• Multimodal Models:

  • Llama 4 Scout / Maverick – Native multimodal models supporting text and image understanding
  • Llama 3.2 Vision – Earlier vision-capable Llama model family, still relevant for image understanding workflows

• Specialized Models:

  • Code Llama (legacy) – Earlier coding-focused models, still used but largely superseded by newer Llama 3.x / 4.x models
  • Instruct / Chat variants – Fine-tuned versions for conversational and task-oriented use

• Open-Weight Strategy:

  • Models are released with weights (not fully open-source, but widely usable)
  • Can be fine-tuned, self-hosted, and deployed without API dependency
  • Strong ecosystem across Hugging Face and local inference tools

• Local & Deployment Ecosystem:

  • Runs locally on consumer and enterprise hardware depending on size
  • Common tooling:
    • Ollama
    • LM Studio
    • vLLM / Text Generation Inference (TGI)
  • Supported across major cloud providers (AWS, Azure, GCP)

• Meta AI Assistant:

  • Meta AI – Integrated assistant across WhatsApp, Instagram, Messenger, and web
  • Powered by latest Llama models and Meta’s internal systems

---

Mistral AI

🔝 Back to Top

Based in Paris 🇪🇺, Mistral AI is a leading European AI company focused on efficient, high-performance models across open-weight and commercial offerings. It has rapidly expanded into a full-stack AI platform including multimodal, reasoning, coding, and audio systems.

• Le Chat - Mistral’s official conversational interface.

• Flagship Models (Current):

  • Mistral Medium 3.5 – Current flagship model for instruction-following, reasoning, coding, long-horizon agents, and enterprise workflows
  • Mistral Large 3 – Open-weight flagship model with multimodal capabilities, mixture-of-experts architecture, and strong multilingual performance
  • Mistral Small 4 – Efficient hybrid model combining instruction-following, reasoning, coding, and multimodal capabilities

• Open / Edge Models (Ministral):

  • Ministral 3 (3B / 8B / 14B) – Lightweight multimodal models designed for local deployment and edge devices
  • Strong focus on running efficiently on consumer hardware, embedded systems, and self-hosted deployments

• Reasoning Models (Magistral):

  • Magistral Medium / Small – Dedicated reasoning models designed for structured problem solving and deeper reasoning workflows

• Coding Models:

  • Codestral – Core code generation and completion model
  • Codestral 25.08 – Updated coding model for enterprise code completion and developer workflows
  • Devstral 2 – Frontier agentic coding model designed for multi-step software engineering tasks

• Audio & Multimodal:

  • Voxtral – Speech understanding and transcription models
  • Voxtral TTS – Text-to-speech model with multilingual support and voice generation capabilities
  • OCR 3 – Document understanding, OCR, and structured extraction model

• Older / Still Used Models:

  • Mixtral (8x7B / 8x22B) – Earlier MoE models, still widely used but no longer flagship
  • Mistral 7B – Legacy lightweight model, now mostly replaced by newer small models

• Platform & Ecosystem:

  • Mistral AI Studio / API – Developer platform for building agents, applications, and model-based workflows
  • Mistral Vibe – Agentic workspace for long-horizon productivity and coding workflows
  • Strong support for:
    • Hugging Face
    • Self-hosted inference (vLLM, TGI, Ollama)
    • Cloud deployment (AWS, Azure, GCP)

• Key Strengths:

  • Efficiency-first engineering – High performance per compute cost
  • Open-weight leadership in Europe – Major alternative to US/China AI providers
  • Full-stack expansion – Now covers text, code, audio, vision, documents, and agents

---

Perplexity

Perplexity is an AI-based search and research platform that combines large language models with real-time web retrieval and source citations.

• Perplexity – Interface for querying, browsing sources, and generating structured answers.

• Models (Sonar family):

  • Sonar – Default model for search-based queries
  • Sonar Pro – Higher-depth retrieval and synthesis
  • Sonar Reasoning Pro – Optimized for multi-step reasoning tasks
  • Sonar Deep Research – Designed for longer, multi-source analysis and report generation
  • And many more from other providers integrated in their system

• Capabilities:

  • Real-time web search with cited sources
  • Multi-step query decomposition and synthesis
  • File and document analysis
  • Structured outputs and summaries
  • Routing across multiple underlying models

• Agent / Computer features:

  • Comet – AI browser with assistant features for browsing, summarizing, email, shopping, and web tasks
  • Computer – Agentic assistant for longer workflows, browser automation, background tasks, and connected apps
  • Supports tool usage and multi-step workflows
  • Includes logging, control, and interruption mechanisms

• API / Platform:

  • Sonar API for web-grounded responses
  • Agent API for model routing, tools, structured outputs, and OpenAI-compatible workflows
  • Search API for ranked web search results
  • Embeddings API for text embeddings
  • Supports streaming, structured outputs, and tool usage

• Access:

  • Free tier with limited usage
  • Pro tier (~$20/month) with higher limits and advanced models
  • Max tier for heavier usage, advanced agent features, and early access
  • Enterprise tiers for teams, admin controls, security, and higher limits

---

🇨🇳 Leading Chinese AI Models

🔝 Back to Top

This section covers several of the most important AI model ecosystems developed by Chinese labs and cloud platforms, including their current flagship models, API pricing, and paid plans where publicly documented.

---

Qwen

Developed by Alibaba Cloud, Qwen spans consumer chat, commercial APIs, open-weight models, coding models, and multimodal systems.

• Qwen Chat – Main interface for interacting with Qwen models.
• Commercial / API models:
  • qwen3.6-max-preview – Highest-end Qwen model for reasoning-heavy and complex agent workloads
  • qwen3.6-plus – Current recommended flagship-style model with 1M context, tool use, coding, and multimodal support
  • qwen3.6-flash – Lower-cost model with near-flagship capabilities and 1M context
  • qwen3.5-plus – Earlier high-capability multimodal model, still available
  • qwen-plus – Balanced production model
  • qwen-flash / qwen3.5-flash – Lower-cost / faster models
• Open-weight models:
  • Qwen3
  • Qwen2.5
• Specialized models:
  • Qwen3-Coder – Coding-focused family
  • Qwen3-VL – Vision-language models
  • Qwen3-Omni – Unified multimodal family
  • Qwen3.5-Omni – Newer multimodal / realtime model line, with preview models temporarily free for model invocation
  • Qwen3-TTS – Speech / voice generation
• API Pricing (USD, official):
  • qwen3.6-max-preview – tiered pricing; starts around $1.20 / 1M input, $6.00 / 1M output for smaller requests
  • qwen3.5-plus – global pricing starts at $0.115 / 1M input, $0.688 / 1M output
  • qwen-plus – EU / Hong Kong deployment starts at $0.40 / 1M input, $1.20 / 1M thinking output, $4.00 / 1M non-thinking output
  • qwen3.5-flash – starts at $0.029 / 1M input, $0.287 / 1M output for requests up to 128K tokens
  • qwen-flash – global pricing starts at $0.05 / 1M input, $0.40 / 1M output up to 256K input
  • qwen3.5-omni-plus / flash – preview model invocation is temporarily free, but tool calling may still incur fees
• Plans / subscriptions:
  • Model Studio Coding Plan Pro – $50/month, includes access to selected coding / agent models including Qwen, Kimi, GLM, and MiniMax models
  • Qwen Code OAuth – Qwen documentation states users can get 1,000 free calls per day to Qwen3.6-Plus through Qwen Code OAuth

---

DeepSeek

DeepSeek focuses on very low-cost reasoning and strong developer adoption, with a simple public API surface and major open-weight releases.

• DeepSeek Platform – API and playground
• Core API models:
  • deepseek-v4-flash – cheaper V4 model, supports thinking and non-thinking modes
  • deepseek-v4-pro – stronger V4 model, now priced much lower after the V4-Pro price cut
  • deepseek-chat – legacy alias currently mapped to DeepSeek-V4-Flash non-thinking mode
  • deepseek-reasoner – legacy alias currently mapped to DeepSeek-V4-Flash thinking mode
• Open / research models:
  • DeepSeek-V3
  • DeepSeek-R1
  • DeepSeek-VL2
  • Janus / Janus-Pro
• API Pricing (USD, official):
  • deepseek-v4-flash:
    • Input cache hit – $0.0028 / 1M tokens
    • Input cache miss – $0.14 / 1M tokens
    • Output – $0.28 / 1M tokens
  • deepseek-v4-pro:
    • Input cache hit – $0.003625 / 1M tokens
    • Input cache miss – $0.435 / 1M tokens
    • Output – $0.87 / 1M tokens
  • deepseek-v4-pro pricing now reflects the major 75% price reduction
• Capabilities:
  • 1M context on current public API endpoints
  • Up to 384K max output
  • Thinking mode, tool calls, JSON output, prefix completion, FIM completion, context caching, and OpenAI / Anthropic-compatible API usage
• Plans / subscriptions:
  • No subscription tiers
  • No tiered pricing plans
  • Usage is prepaid / top-up based
• Positioning:
  • One of the cheapest serious APIs on the market
  • Especially attractive for reasoning, coding, and cost-sensitive agent workflows
  • Current V4 pricing is aggressive enough that DeepSeek is now one of the main pressure points in global API pricing

---

Kimi

Kimi (Moonshot AI) focuses on long-context work, coding, multimodal input, and agent-based workflows.

• Kimi – Main consumer interface
• Kimi API Platform – Developer platform
• Current models:
  • kimi-k2.6 – latest Kimi model with stronger long-context coding and agent stability
  • kimi-k2.5 – flagship multimodal model with vision, thinking / non-thinking modes, and agent support
  • moonshot-v1 – older general-purpose family
• Older / legacy models:
  • kimi-k2-0905-preview
  • kimi-k2-turbo-preview
  • kimi-k2-thinking
  • These are now less central than K2.5 / K2.6 for current usage
• API Pricing (USD, official):
  • kimi-k2.6 – $0.16 / 1M input (cache hit), $0.95 / 1M input (cache miss), $4.00 / 1M output
  • kimi-k2.5 – $0.10 / 1M input (cache hit), $0.60 / 1M input (cache miss), $3.00 / 1M output
  • moonshot-v1-8k – $0.20 / 1M input, $2.00 / 1M output
  • moonshot-v1-32k – $1.00 / 1M input, $3.00 / 1M output
  • moonshot-v1-128k – $2.00 / 1M input, $5.00 / 1M output
  • Web search tool – $0.005 per successful tool call
  • Batch API – cheaper batch pricing available for K2.5 / K2.6
• Capabilities:
  • 256K context on K2.5 / K2.6
  • Tool calls, JSON mode, automatic context caching, internet search, multimodal input, and agent-style workflows
• Plans / subscriptions:
  • Kimi Membership – official paid membership page with multiple tiers
  • Kimi Code is a membership benefit; its docs state quota is shared with Kimi membership
  • Kimi Business exists for enterprise users via contact sales
  • API access is recharge-based; minimum recharge is $1
  • When cumulative recharge reaches $5, Kimi currently grants a $5 voucher
• Positioning:
  • Strong for long-context work, coding, multimodal tasks, and agent workflows
  • More productized than many rival model providers
  • K2.6 makes Kimi more relevant again for coding and long-running agent workflows

---

GLM / Z.AI / ChatGLM

GLM is Zhipu / Z.AI’s model family, aimed at reasoning, coding, agents, and enterprise deployments.

• ChatGLM – Main consumer interface
• Z.AI – International-facing GLM assistant and developer ecosystem
• BigModel – China API / platform
• Z.AI Developer Platform – International API documentation
• Current models:
  • GLM-5.1 – current flagship model for agentic engineering and stronger coding workflows
  • GLM-5 – open-weight flagship foundation model for long-horizon agentic tasks
  • GLM-5-Turbo – faster commercial API variant
  • GLM-5V-Turbo – vision-capable Turbo variant
  • GLM-4.7 / GLM-4.7-FlashX – cheaper previous-generation production models
  • GLM-4.7-Flash / GLM-4.5-Flash – free API models on Z.AI’s pricing page
• API Pricing (USD, official Z.AI pricing):
  • GLM-5.1 – $1.40 / 1M input, $0.26 / 1M cached input, $4.40 / 1M output
  • GLM-5 – $1.00 / 1M input, $0.20 / 1M cached input, $3.20 / 1M output
  • GLM-5-Turbo – $1.20 / 1M input, $0.24 / 1M cached input, $4.00 / 1M output
  • GLM-4.7 – $0.60 / 1M input, $0.11 / 1M cached input, $2.20 / 1M output
  • GLM-4.7-FlashX – $0.07 / 1M input, $0.01 / 1M cached input, $0.40 / 1M output
  • GLM-4.7-Flash – Free
  • GLM-4.5-Flash – Free
• API Pricing (official BigModel China pricing, CNY; approx USD varies by exchange rate):
  • GLM-5.1 – starts at ¥6 / 1M input, ¥24 / 1M output
  • GLM-5 – starts at ¥4 / 1M input, ¥18 / 1M output
  • GLM-5-Turbo – starts at ¥5 / 1M input, ¥22 / 1M output
• Capabilities:
  • Around 200K context on GLM-5 class models
  • Tool calling, structured output, thinking modes, context caching, coding, and long-horizon agent execution
  • GLM-5 / 5.1 weights are available through Hugging Face / ModelScope
• Plans / subscriptions:
  • GLM Coding Plan – subscription access for coding tools and agent workflows
  • Supports usage through tools such as Claude Code-style setups, Cline, Kilo Code, OpenCode, Cursor-style workflows, and MCP integrations
• Positioning:
  • One of the strongest Chinese open-weight coding / agent model families
  • Very aggressive pricing at the low end due to free Flash models and cheap FlashX pricing
  • Particularly relevant for agentic coding and long-running development tasks

---

MiniMax

MiniMax develops multimodal foundation models with a strong focus on coding, agents, long context, audio, video, and consumer products such as Hailuo.

• MiniMax – Main company site
• MiniMax API Platform – Developer platform
• Current models:
  • MiniMax-M2.5 – Current flagship text model for coding, agents, search, tool use, and office workflows
  • MiniMax-M2.5-highspeed – Faster variant with the same core capability profile
  • MiniMax-M2.1 / M2 – Earlier coding / agent models
  • Hailuo – MiniMax video generation product family
• API Pricing (USD, official):
  • MiniMax-M2.5 – $0.30 / 1M input, $1.20 / 1M output, $0.03 / 1M prompt-cache read, $0.375 / 1M prompt-cache write
  • MiniMax-M2.5-highspeed – $0.60 / 1M input, $2.40 / 1M output, $0.03 / 1M prompt-cache read, $0.375 / 1M prompt-cache write
• Plans / subscriptions:
  • Plus – $20/month
  • Max – $50/month
  • Ultra – $120/month
  • Token plans cover eligible API Platform resources and are aimed at agent, coding, and multimodal workflows
• Capabilities:
  • Long-context coding and agent workflows
  • Tool use, prompt caching, search-oriented workflows, and high-throughput coding tasks
• Positioning:
  • Extremely cost-aggressive for coding and agent workflows
  • One of the Chinese providers pushing “frontier-ish” model usage toward very low per-task costs

---

Baidu ERNIE

Baidu’s ERNIE family powers Baidu’s consumer assistant and Qianfan enterprise platform, with strong Chinese-language, multimodal, and reasoning-focused model lines.

• ERNIE – Main consumer / model site
• Baidu Qianfan – Enterprise API / model platform
• Current models:
  • ERNIE 5.1 – Newer ERNIE generation focused on lower inference cost and strong reasoning / writing performance
  • ERNIE 5.0 – Native omni-modal foundation model
  • ERNIE X1 / X1.1 – Reasoning-focused model line
  • ERNIE 4.5 – Earlier open / production model family
  • ERNIE-Image – Open image generation model line
• API Pricing (USD, official Qianfan international page where listed):
  • ERNIE 5.0 – $1.40 / 1M input, $5.60 / 1M output
  • DeepSeek-V3.2 via Qianfan – $0.50 / 1M input, $0.50 / 1M cache hit, $1.50 / 1M output
• Capabilities:
  • Text generation, deep thinking, visual understanding, OCR, structured output, function calling, and dialogue continuation
• Positioning:
  • Major Chinese incumbent model ecosystem
  • Strong domestic platform integration through Baidu Cloud / Qianfan
  • Less frictionless internationally than Qwen, DeepSeek, Kimi, GLM, or MiniMax, but important enough to track

---

Hunyuan

Hunyuan is Tencent’s model platform, spanning text generation, reasoning, translation, OCR, 3D, video, world models, and broader multimodal tooling.

• Hunyuan – Main product site
• Tencent HY / Hunyuan docs – API overview

…view the full README on GitHub.