Models & infrastructure

LM Studio lets you run large language models directly on your own computer without an internet connection, keeping your data private. It's aimed at researchers and developers who want full control over their data.

Arena AI lets people send one prompt to two anonymous AI models, vote for the better answer, and the votes build a public ranking. It helps anyone see which AI models perform best.

Artificial Analysis independently benchmarks AI models on quality, speed, and cost, with recommendations for different needs. It helps developers pick the right model to use.

DeepSeek-V3 is an open-source AI model with strong performance in coding, math and reasoning, offering advanced capability at low cost.

Vercel AI Playground lets users test and compare multiple AI models from providers like OpenAI, Anthropic and Cohere in one place. It's useful for developers and researchers deciding which model fits their project.

PromptPerfect refines your prompts for large language models in seconds, so you get better results without endless trial and error.

Cohere provides a language AI platform for businesses to build products with content generation, summarization, semantic search, and classification in 100+ languages.

TrainEngine.ai (Dreambooth) helps you train and customize your own AI models using images, audio, and other data, at different accuracy levels.

Ailiverse NeuCore is an API for training computer vision models with less data and faster GPU training. It supports image classification, object detection, action recognition, and deepfake detection.

LandingLens is a platform for building computer vision models without deep AI experience. Users can upload images, label objects, train a model, and deploy it.

AssemblyAI is a developer API platform that converts speech to text, both live and from recordings. It also provides summaries, entity extraction, and sentiment analysis on conversations.

Yonoo brings major AI models like ChatGPT, Claude, and Gemini together and routes your question to whichever fits best. Users get text, images, and video without switching between separate apps.

Metatext lets anyone build a language AI model without needing coding or machine-learning expertise, just by uploading their own data. It suits businesses that want custom AI without hiring specialists.

A no-code interface for fine-tuning large language models to fit specific needs. It helps researchers and professionals customize AI models without deep coding skills.

A cloud platform offering GPU computing power for training or running AI models. Users pay based on usage time, making it useful for data professionals and AI developers.

An app that lets Mac users run open-source AI models directly on their device without sending data online. It helps privacy-conscious users benefit from AI without worrying about data leaks.

Lightning AI shortens the path from training a model to running it in front of users, using ready-made parts, with an eye on keeping the cloud bill down.

A web-based framework that lets developers build and deploy AI products by combining different language models. Good for developers wanting to build AI solutions without heavy infrastructure.

Agent.so measures how well a machine learning model performs, including accuracy and error rates. It helps developers understand their model's behavior before deploying it.

Quartzite AI is a workspace for writing and refining complex prompts for AI models like GPT-4. It helps technical teams track prompt versions and reuse templates.

Autoblocks AI helps teams build, deploy, and monitor large language models at scale, making it easier to ensure AI systems run reliably.

Pinecone is a specialized database that helps developers build fast AI-powered search into their applications, designed to be secure and scale with growing usage.

PromptBlaze is a marketplace and browser extension for storing, buying, and selling AI prompts for tools like Claude and ChatGPT. Users can chain several prompts together into automated workflows.

AI-Flow is a free, open-source app that connects multiple AI models together to give varied perspectives on one prompt. All data stays on your own computer, which suits privacy-conscious users.

astica is a demo of AI vision and voice services that describe images, detect faces, transcribe audio, and flag inappropriate content. It suits developers testing computer-vision APIs before integrating them.

VERSES builds KOSM, an operating system that links data, policies, and simulations across systems into a shared distributed intelligence. It also offers WAYFINDER for optimizing warehouse operations.

Scade.pro is a no-code platform that lets users integrate over 1,500 AI models for audio, image, text, and video into their projects to speed up work and cut costs.

AI/ML API offers a single interface to access over 100 AI models for chat, image generation, coding and more. Developers can switch easily from OpenAI with minimal code changes, saving cost and integration effort.

Beam lets developers run AI projects without managing infrastructure, using serverless APIs and GPUs with just a few lines of code. Users pay only for what they use.

Mem0 is a developer tool that lets AI applications remember past conversations and user preferences even after a session ends.

Synexa AI lets developers run AI models to generate images, video, and other media through cloud APIs and playgrounds. It helps businesses add AI-generated content to their apps quickly and affordably.

Google AI Edge lets developers run AI models directly on phones and other devices, working even offline while keeping user data private.

RunComfy sets up ComfyUI environments in the cloud without technical hassle, so creators can generate AI images and videos without worrying about configuration.

CometAPI gives developers one connection to over 500 AI models covering text, images, video and music. Instead of hooking up to each provider separately, you go through a single API.

Metal is a managed storage platform that makes it easy to work with embeddings in your applications. Developers use it to plug into various vector databases without building the infrastructure themselves.

Snow Goose gives you one place to use major AI models like GPT and Claude without switching apps, for a fixed monthly price. It suits writers, researchers and developers.

BrewPrompts helps you write better AI prompts, with a builder for text prompts and one for art prompts, plus a place to save favorites for reuse. Designers and marketers use it to save time and get more reliable results.

Cerebrium is a framework for training, deploying, and monitoring machine learning models with minimal code, aimed at developers who want scalable AI infrastructure without the hassle.

GMTech lets you compare several AI chat models and image generators side by side in one app. It helps researchers, developers, and creators quickly find the AI tool that best suits their needs.

Testmyprompt helps AI automation agencies build, refine and test their prompts faster and more consistently. It cuts down on errors and raises the quality of AI responses.

Empromptu is a no-code platform for building complete AI applications, bundling models, retrieval-augmented generation, and infrastructure together. Companies use it to get production-ready solutions running quickly, either on their own servers or in the cloud.

Nailedit.ai lets you compare answers from ChatGPT, Claude, and Gemini side by side for the same prompt. It helps writers and researchers pick the best response without switching apps.

Automorphic offers tools to fine-tune language models with small samples, quickly load model adapters, and a firewall called Aegis that detects prompt injection and data leaks.

LLM List is a directory comparing various large language models, helping developers and businesses find and choose the right model for their needs.

Keywords AI simplifies building and managing LLM applications by offering one API to access over 200 models, plus monitoring and analytics. It helps developers and AI startups improve reliability and performance.

Memara gives AI systems like Claude and ChatGPT the ability to remember facts and preferences across separate conversations, instead of starting fresh each time.

ModelsLab provides cloud-based APIs for generating images, video, audio and text with AI, letting developers add these capabilities to their apps without building their own infrastructure.

Magic Potion is a visual editor for building, testing, and organizing prompts across models like OpenAI, Anthropic, and Google, with live responses as you work.

Memory Plugin is a browser extension that gives ChatGPT, Claude, and Gemini long-term memory of your conversations. You can organize memories into groups and control what's kept.

NodeAI lets people tap into GPU computing power from around the world on a pay-as-you-go basis. GPU owners can rent out their hardware for extra income.

Plurality Network stores a user's memories and preferences securely so they can carry over across different AI apps and models without starting from scratch each time.

Promptve helps teams create, test and store AI prompts in one place, maintaining consistency and tracking how well each prompt performs.

Sieve is a cloud platform for building and running AI apps that combine multiple models, like understanding audio or generating video. Developers value how easy it scales.

Featherless.ai gives API access to a wide range of text-generation models, like LLaMA and Mistral, without you managing your own servers. It suits developers who want flexible, affordable access to many models.

Vellum is a developer platform for building, testing and deploying AI-powered applications, with tools for prompt iteration and performance monitoring in production. It works across all the major AI providers.

Zep is a memory layer for AI assistants that lets them remember past conversations and get more accurate over time. Developers use it to cut down hallucinations and build smarter AI applications.

CanIRun.ai scans your computer's GPU, CPU and RAM and checks it against hundreds of AI models' requirements, so you know which ones you can run locally. It suits anyone deciding whether to upgrade their machine for AI.

CloudGo.ai analyzes your cloud setup and code to give quick recommendations on cost, security, and architecture, helping tech teams make better infrastructure decisions and avoid manual configuration errors.

Kolena is a platform for testing and improving machine learning models, helping AI engineers ensure quality and explainability before deployment.

Mirai is a developer SDK that lets AI models run directly on iPhone or Mac devices offline, using Apple hardware for faster processing while keeping user data private on-device.

Toolhouse is a platform for managing and optimizing the tools that large language models use. It helps developers plug in tools quickly, keep latency low, and debug what's happening under the hood.

Athina AI monitors AI systems in production and catches when they hallucinate or produce low-quality answers, helping companies fix issues before they cause harm. It also lets teams compare different models and prompts.

Confident AI is an open-source platform for testing AI models - comparing versions, catching errors and making sure an AI system's answers can be trusted.

Gentrace helps evaluate and monitor AI model performance in production, offering strong security controls and self-hosting options.

Kolosal AI is an open-source app that lets you download and run large language models directly on your own device, keeping your data off the cloud.

Together AI is a cloud platform for building, deploying and scaling AI models, offering fast, scalable infrastructure for companies and researchers.

WhatLLM.org compares over 100 AI models on price, speed and quality using one standardized index, updated daily. It helps developers pick the best model for tasks like coding or reasoning.

Klavis AI is an open-source platform that connects AI agents to thousands of external tools, removing the hassle of wiring up authentication for each one separately.

Letta is a developer platform for building AI agents with lasting memory, ones that remember past conversations and can explain their reasoning.

MemMachine is an open-source memory layer for AI agents, letting them recall user preferences and past context instead of starting fresh every conversation.

nexos.ai brings multiple AI models together behind one company gateway, letting teams compare outputs, manage costs, and set security policies in one place.

SEAL Leaderboard evaluates and ranks large AI models like GPT, Claude and Gemini using curated datasets and real-world usage, so people can see which model performs best.

EverMind is an open-source memory platform that gives AI models long-term memory and a persistent identity, so agents can recall past conversations and stay consistent over time.

Fiddler AI monitors and explains how machine learning models behave, detecting issues like bias or performance drift. It helps data and security teams ensure their AI systems remain transparent and trustworthy.

Kento cuts AI usage costs by up to 40% by caching answers to duplicate or similar prompts and serving them instantly instead of paying the model again, with just one line of code to integrate.

MemU is an AI memory layer for LLM-based applications, especially AI companions - it autonomously organizes, links and refines stored information into a growing knowledge graph over time.

PromptPanda centralizes and tags all of a team's AI prompts in one place, scoring their quality and suggesting improvements so everyone's AI output stays consistent with the brand voice.

HasMCP turns your API documentation (OpenAPI/Swagger) into ready-to-use tools for AI agents with almost no extra code, while handling authentication and secrets. It suits developers who want their agents to talk to backend services without building glue code.

Hebbrix gives AI agents long-term, searchable memory, combining several search methods with a knowledge graph so responses stay grounded in context. It suits developers building AI apps that need to remember past conversations accurately.

IQuest Coder is an open-source, code-focused language model that can understand a whole repository's history, review its own output, and generate, test, and fix multi-file code at once, with a 128K context window. It suits developers who want a self-hostable coding model.

KeyAPI gives developers one key to pull profiles, posts, videos, and analytics from over 20 social platforms, formatted for AI systems to use directly. It suits teams building agents or models that need social media data at scale.

MemoryLake carries your context and history between ChatGPT, Claude, Gemini, and other AI systems, storing it securely so you don't have to re-explain yourself every conversation. It suits people who switch between multiple AI tools often.

OneRouter is a single API gateway that lets companies access and switch between more than 100 AI models, with automatic routing, caching, and built-in cost and safety controls. It suits larger organizations that want to manage all their AI usage from one place.

Respan monitors AI systems in production, tracing every request, scoring answer quality, and controlling costs across more than 500 models through one gateway. It suits developers who need to keep agents and chatbots reliable and on budget.

Talkory.ai queries GPT, Claude, Gemini, Grok and Perplexity all at once, shows their answers side by side, and synthesizes a ranked consensus answer with a confidence score, cutting down on hallucinations and tab-switching.

Clawputer (OpenComputer) lets developers deploy AI agents as TypeScript functions running on full, isolated Linux virtual machines, with shell, filesystem, and network access, and sessions that persist beyond a single request. It suits developers needing a real sandboxed machine.

HelpMyAgent offers a catalog of pay-per-call APIs built specifically for AI agents, with transparent pricing, documented data sources, and real sample responses before you integrate. It suits developers who need specific data without a monthly subscription.

LLMWise sends one prompt to many AI models at once -- GPT, Claude, Gemini and more -- then lets you compare, blend, or auto-pick the best answer.

Osaurus is free, open-source Mac software that runs AI models directly on your device with no data leaving it. You can optionally connect ChatGPT, Claude, or Gemini when you need extra power, and build agents that act on their own.

PromptUnit is a middle layer that automatically routes each request to the cheapest AI model that still meets your quality bar, and compresses prompts to cut costs further. You only pay a share of the savings it actually delivers.

Abliteration is an AI model API with fewer of the usual safety refusals, built for cybersecurity and high-risk research teams, with policy rules that check every request before it's answered.

AICanRun tells you which local AI models your phone can actually run offline, based on its chip and memory, plus the speed you can expect.

Alma is a memory engine that gives conversational AI the ability to recall past interactions and a user's preferences, instead of starting fresh every time. It helps developers build assistants that get more personalized the more they are used.

AnyAPI gives developers one key to reach major AI models like GPT, Claude, Mistral, and Gemini, with easy prompt testing and quick switching between them.

BenchGen tests AI agents inside simulated environments to see how they really perform, then turns the results into training data for further improvement.

Cycles sets spending limits for AI agents before they act - it estimates and reserves a budget for each request, then allows or blocks the action based on rules you set. It helps developers stop AI costs from spiraling out of control.

Dial gives AI agents real phone numbers through one API, so they can make calls, send texts, and iMessage -- with live transcription on every conversation.

EUrouter provides a single API gateway to over 100 AI models while keeping all data inside the EU under GDPR rules, automatically picking the best model for cost or speed.

Inference.net is a cloud service that lets engineering teams run and manage AI models at scale with lower cost and latency. It also tracks performance so teams can spot and fix problems in their models.

Maximem stores your personal and work context privately and shares it across AI tools like ChatGPT, Claude, and Gemini, so you stop re-explaining your background every time.

Mireye gives AI agents real-world location data -- flood zones, elevation, soil type, and more -- through one API, with every answer citing its source.

Nativ is a Mac app that runs open-source AI models straight on your computer, with no account or cloud service needed.

Observyze helps developers track every call their AI model makes, measure token costs in real time, and catch problems like hallucinations or prompt injection attacks.

OpenServ provides AI agent infrastructure with built-in validation and audit trails, built for organizations like banks and governments that need transparency in AI decisions.

OrcaRouter takes requests from your app and sends each one to whichever of 200-plus AI models fits best on cost, speed, and quality. It helps developers cut AI spending without sacrificing reliability.

Oumi helps teams build, train, and deploy their own specialized AI models just from a plain description, letting them own the training data and model weights instead of depending on a big provider.

Buy once and run open-source AI models like Llama or Qwen entirely on your iPhone, iPad, or Mac, working offline after the first download, with tools built into Siri and Shortcuts.

rightmodeler is a command-line tool that tests cheaper AI models against previously accepted outputs, helping teams switch models without losing quality.

SignalBrain-OS runs multi-agent AI decisions locally on the machine, producing extremely fast, auditable, tamper-evident records of every action for high-stakes work.

Syzygy runs an open AI model entirely on your Mac, giving it access to your files and apps as a work assistant without sending anything to the cloud.

Tangle is an open-source drag-and-drop editor for building machine learning pipelines, letting you wrap existing code in any language into connected steps.

Tinker is a playground for the Inkling reasoning model, where you can adjust how hard it thinks, turn on web search, and test prompts without your chats being saved.

Vectoralix turns your files, code, and tools into a live MCP endpoint that AI agents can call, with document search and safe code execution built in. It helps developers connect their own knowledge to AI agents without building the infrastructure themselves.

Vivgrid is infrastructure for building and running multi-agent AI systems, tracing every step, enforcing safety rules, and delivering responses with very low latency.

AIxBlock provides speech and language data services for training AI models, including collection, transcription, labeling, and RLHF data. Its self-hostable platform offers a training pipeline, dataset tools, and access to a GPU marketplace, aimed at AI development teams.

APIPark is an open-source gateway and developer portal for handling LLM traffic in production. It exposes an OpenAI-compatible interface so you can plug in different models without touching your existing app code, while adding load balancing, quotas, monitoring, and access control.

Bakery is an open-source platform that lets people fine-tune and sell their own AI models with a single click. It offers tools for uploading datasets, adjusting model settings, and listing models on its marketplace, aimed at AI engineers and researchers.

This platform labels images, video, audio, and other data to train AI models, with managed annotation services available. It suits companies building their own AI systems.

Citations is a Claude API feature that grounds answers in documents you supply, pointing to the exact sentences used so apps can show where an answer came from.

Clarifai is a platform for building, training, and deploying AI models for computer vision, text, and audio, including tools for labeling data.

Galactica is a large language model built by Meta for the research community, released openly with its paper, code, and demo. It acknowledges its own limits, including the risk of producing inaccurate answers.

IBM watsonx is an enterprise platform for building, testing, and deploying trustworthy AI models while managing company data at scale.

This is Meituan's open-source large language model, built for fast responses, tool use, and multi-step reasoning. Developers can access it free via Hugging Face and GitHub.

Mistral AI supplies its own models plus the place to build, test and run things with them: Studio for agents and apps, Forge for training, and hosting to keep them running.

Ollama lets you download and run open language models directly on your own computer, without needing the internet every time. It offers a command-line interface, model management, and an OpenAI-compatible API.