Best Free AI Courses From Anthropic, OpenAI, Google, Microsoft, NVIDIA and Hugging Face in 2027

A laptop showing a course video on a desk in the aisle of a dim server hall

The free AI courses Anthropic, OpenAI, Google, Microsoft, NVIDIA and Hugging Face offer in 2027, compared by audience, format, certificate and vendor tilt.

AI / AI Education / Free AI courses

Every major AI lab now runs a free academy, and the best of them are worth more than many paid courses. Anthropic Academy offers free courses with certificates on Claude, Claude Code, the Claude API and the Model Context Protocol; OpenAI Academy covers ChatGPT, Codex and the OpenAI API; Microsoft publishes MIT-licensed curricula on generative AI, agents and MCP; Google offers its Machine Learning Crash Course, a generative AI path and, with Kaggle, a five-day agents intensive; NVIDIA makes many Deep Learning Institute courses free; and Hugging Face runs the most vendor-neutral courses of all. We checked each provider's own pages on 19 November 2026. Each lab teaches its own stack first, so the useful question is which lab's tilt matches your work.

Key takeaways

  1. Hugging Face is the most neutral lab course provider, teaching several frameworks and coding agents side by side.
  2. Anthropic Academy and the Hugging Face Agents Course issue free certificates; Google awards skill badges, and NVIDIA certifies only select courses.
  3. Microsoft's open curricula on GitHub are the easiest to adapt for internal training, thanks to the MIT license.
  4. Pick the academy that matches the platform you ship on, then add one neutral course to avoid learning only one vendor's abstractions.

Which free courses from AI labs are worth taking?

Ranked by how much a practitioner or technology leader gets out of each one. "Checked" names the source behind each number and when we looked.

Anthropic

  1. Free · certificates of completion · AI Fluency, Claude 101, Claude Code, Claude API, MCP, agent skills, subagents, Claude Enterprise rollout

    For: Teams on Claude, developers writing MCP servers, and staff who need AI fluency training.

    Anthropic's catalog spans the widest range of audiences of any lab: AI Fluency editions for educators, nonprofits and small businesses, developer courses on the API and MCP, and a course on the decisions behind a Claude Enterprise rollout. The MCP courses teach a protocol that other vendors' clients support too. Courses are free and certify on completion.

    Checked: Course catalog (courses marked Free), 19 Nov 2026

OpenAI

  1. Free sign-up · courses for workers, builders, leaders and educators · regular livestreams and in-person workshops

    For: Organizations on ChatGPT, and developers adopting Codex or the OpenAI API.

    OpenAI's strength is its live program: livestreams for specific professions, in-person workshops in several cities and Codex bootcamps for engineers on 3 and 10 Dec 2026. The on-demand courses are organized into four tracks, from applying AI at work to leading adoption. It is open to anyone with a free account.

    Checked: Provider site (home and events), 19 Nov 2026

Google

  1. Open without sign-in · self-paced modules with videos, visualizations and exercises

    For: Developers and analysts who need machine learning fundamentals before generative AI.

    Google says millions of people have used the course since 2018, and the refreshed version adds embeddings, an introduction to LLMs and production ML. It is the least vendor-specific Google course, since the fundamentals apply anywhere. Modules are self-contained, so experienced learners can jump in.

    Checked: Google for Developers course page, 19 Nov 2026

  2. Free · 4 activities · 45-minute opening course · skill badge

    For: Non-specialists in Google Cloud organizations who need a credentialed generative AI overview.

    The path covers generative AI concepts, large language models and responsible AI, and the opening course points to Google's tools for building generative AI apps. The badge is useful inside Google Cloud partner and customer organizations. It is short enough to complete in a lunch break.

    Checked: Google Skills path and course pages, 19 Nov 2026

  3. Free · self-paced Learn Guide · 1.5M+ learners in the live run

    For: Developers who want Google's view of agent design, with runnable notebooks.

    Google and Kaggle turned the live intensive's whitepapers, notebooks and capstone into a permanent Learn Guide after more than 1.5 million people joined. It is the most compact free agent course from a lab. Google said a new live intensive was planned for 2026.

    Checked: Kaggle Learn Guide page (title) and Google blog recap, 19 Nov 2026

Microsoft

  1. Free, MIT license · 21 lessons · 124k GitHub stars

    For: Developers on Azure, and L&D teams that want a course they can fork and adapt.

    Microsoft's Cloud Advocates wrote it, and its MIT license lets a company copy and customize it for internal training, which no other lab course allows as clearly. Lessons support Azure OpenAI, Microsoft Foundry Models and the OpenAI API. It has more GitHub stars than any other course repository on this page.

    Checked: GitHub repository page, 19 Nov 2026

  2. Free, MIT license · 18 agent lessons (80k stars) plus an MCP curriculum in six languages (17.7k stars)

    For: Developers building agents on Microsoft Agent Framework, or MCP servers in .NET, Java, TypeScript, Rust or Python.

    The agents course covers design patterns through memory, browser use and agent security, and the MCP curriculum is the only lab course teaching MCP in six languages. Both lean on Microsoft Foundry and Agent Framework. They are the natural next step after Generative AI for Beginners.

    Checked: GitHub repository pages, 19 Nov 2026

NVIDIA

  1. Many self-paced courses free · beginner level, a day or less · certificates in select courses

    For: Engineers who want GPU-backed labs, and teams working with NVIDIA's models and inference stack.

    NVIDIA keeps a set of its top-rated self-paced courses free, designed for beginners and finishable in a day. The AI Learning Essentials hub adds free material on building agents and agentic RAG with Nemotron models. Paid courses and instructor-led workshops sit behind them for teams that need depth.

    Checked: NVIDIA self-paced training and AI Learning Essentials pages, 19 Nov 2026

Hugging Face

  1. Free, with two free certificates · smolagents, LlamaIndex, LangGraph

    For: Developers who want a lab course that does not tie them to one model provider.

    Hugging Face teaches its own smolagents next to LlamaIndex and LangGraph, which makes it the least partisan agent course from any lab. Both certificates are free and have no deadline. Hands-on sections run in preconfigured Hugging Face Spaces.

    Checked: Course introduction page, 19 Nov 2026

  2. Free · 6 units · Claude Code, Codex or OpenCode

    For: Developers who use coding agents from more than one vendor.

    A lab course about other labs' products: it teaches skills, MCP, plugins, subagents and hooks in a way that works across Claude Code, Codex and OpenCode. That cross-vendor framing is rare and useful for teams that have not standardized on one agent.

    Checked: Course welcome page, 19 Nov 2026

  3. Free, no ads · 12 chapters · open models, fine-tuning, reasoning models

    For: Engineers working with open-weight models.

    The only lab course here centered on open models: loading, fine-tuning, sharing and building reasoning models with the Transformers ecosystem. It is completely free and ad-free. It needs good Python and works best after an introductory deep learning course.

    Checked: Course introduction page, 19 Nov 2026

How do the AI labs' free courses differ?

By audience and by openness. Anthropic and OpenAI build academies for whole organizations, with tracks for non-technical staff, managers and developers, and both make sign-up free. Microsoft publishes code-first curricula on GitHub under the MIT license, which means a company can fork and adapt them. Google splits its material across Google for Developers, Google Skills and Kaggle. NVIDIA keeps a free tier of beginner courses inside a mostly paid institute. Hugging Face teaches the open ecosystem and other vendors' tools alongside its own.

Certificates differ too. Anthropic Academy issues certificates of completion on free courses, and Hugging Face's Agents Course issues two free certificates. Google awards skill badges, NVIDIA certifies only select courses, and Microsoft's GitHub curricula and OpenAI's on-demand courses are about learning rather than credentials.

Format is the third difference. OpenAI leans on live events, from profession-specific livestreams to in-person workshops; Kaggle and Google ran their agents course as a five-day live intensive before turning it into a self-paced guide; Anthropic, Microsoft and Hugging Face are almost entirely self-paced. Learners who need a schedule to finish should favor the live formats and catch the recordings for anything they miss.

Should you worry about vendor bias in AI lab courses?

A little, and it is easy to correct for. Each lab's examples use its own models and platform, so a course can teach a vendor's abstractions as if they were the only way to build. The underlying ideas, such as tool calling, retrieval, context management and evaluation, carry over between providers.

The practical answer is to take the academy for the platform you ship on and one neutral course next to it. Hugging Face's Agents and Context courses are the most neutral options, and Anthropic's MCP courses teach an open protocol that clients from several vendors support. For a broader view of how the platforms compare in production, our comparison of Codex, Claude Code and Gemini CLI and our LLM API pricing comparison help with the decision the courses do not make for you.

Which lab academy should a team start with?

Start with the vendor whose tools your team already pays for, because that is where training turns into daily use fastest. Claude teams should begin with Anthropic's Claude 101 and Claude Code in Action; ChatGPT and Codex teams with OpenAI Academy; Azure teams with Microsoft's Generative AI for Beginners. Teams that run open models or several providers should start with Hugging Face.

Then widen the view. A developer who has finished one lab's agent course learns more from a second lab's than from repeating the first, because the differences expose the design choices. Our free AI courses for developers list sets out a full order that mixes the labs.

Listed here, or should be?

If an entry is out of date or something belongs on this list, email [email protected]. Listings are editorial and free. A featured, labeled Sponsored slot is available on request.

Related questions

Does Anthropic offer free AI courses?

Yes. Anthropic Academy offers free courses with certificates of completion, including AI Fluency: Framework & Foundations, Claude 101, Claude Code in Action, Building with the Claude API, two Model Context Protocol courses and courses on agent skills and subagents.

Is OpenAI Academy free?

Yes. OpenAI Academy says sign-up is free and open to all, and it gives access to on-demand courses, livestreams and in-person events, including Codex bootcamp livestreams for developers.

Which AI lab course is the least tied to one vendor?

Hugging Face's courses. The Agents Course teaches smolagents, LlamaIndex and LangGraph, and the Context Course works with Claude Code, Codex and OpenCode.

Are NVIDIA's AI courses free?

Many are. NVIDIA says a number of its popular self-paced Deep Learning Institute courses are free, designed for beginners and can be finished in a day or less. Other self-paced courses and instructor-led workshops are paid, and certificates are offered in select courses.

Explore More

Ready to Find the Right AI Tools?

Browse our data-driven rankings to find the best AI tools for your team.