Prompt Engineering with Llama 2 & 3
by Amit Sangani · DeepLearning.AI
Our Verdict
Worth it — with caveatsDeepLearning.AI's 'Prompt Engineering with Llama 2 & 3' is a strong, free one-afternoon primer on Meta-specific prompting, but only worth taking if Llama is your target stack and you confirm it is currently live. Taught by Amit Sangani, Senior Director of AI Partner Engineering at Meta, it is the most authoritative free introduction to the Llama prompting workflow. Across roughly 1 hour 53 minutes (10 short video lessons, 7 code-along Jupyter notebooks, and one graded quiz) it walks through API calls, multi-turn conversations, few-shot and chain-of-thought prompting, Code Llama for code generation, and the Llama Guard safety model. It earns a 4.4 / 5 rating on Class Central and Coursera (54 reviews) and forum learners call it 'very well done,' but it is genuinely beginner-level and narrow in scope. Important caveat: the course was pulled offline for maintenance in late 2024 due to API deprecation issues outside DeepLearning.AI's control, so availability and code examples may be unstable, and it teaches Llama 2/3 text prompting only, not vision/multimodal despite this catalog id implying 'vision.'
It is a solid, free, official one-afternoon primer for people who already know basic Python/LLMs and want Llama-specific prompting habits. But it is short, beginner-only, tied to a specific (now partly deprecated) hosted API, and offers no broad prompt-engineering theory, so it is worth taking only if the Llama family is your target stack and you can confirm the course is currently live.
Best for: Developers, data scientists, and technical learners who already understand basic Python and the idea of calling an LLM API, and who specifically want to learn how to prompt and choose among Meta's open Llama 2 and Llama 3 models, use Code Llama as a pair-programming helper, and add Llama Guard safety checks. DeepLearning.AI's own 'Who should join' note frames it broadly as anyone interested in prompt engineering who wants to try the Llama models.
Skip if: People who want a general, model-agnostic prompt-engineering foundation (the ChatGPT/OpenAI-focused courses cover that better), anyone seeking depth, fine-tuning, RAG, or agent-building, complete non-coders who do not want to touch Jupyter notebooks, and learners who need vision/multimodal Llama content, which this course does not cover despite the catalog id.
About This Course
Apply prompt engineering techniques to Meta's Llama models for text generation, code generation, and reasoning tasks.
What You'll Learn
Curriculum
Video, ~4 min: course overview and goals.
Video, ~4 min: the Llama 2 & 3 model family and where each fits.
Video with code example, ~16 min: first API calls to the models.
Video with code example, ~11 min: maintaining context across turns and prompt formatting.
Video with code example, ~22 min: few-shot prompting for sentiment and chain-of-thought for reasoning.
Video with code example, ~16 min: how output differs across model variants for the same task.
Video with code example, ~17 min: using Code Llama to write and improve code.
Video with code example, ~15 min: checking prompts and responses for harmful content with the Llama Guard safety model.
Video with code example, ~3 min: optional walkthrough of the helper used to call the API.
Video, ~2 min: wrap-up.
Graded quiz, ~10 min: assessment; the graded item / accomplishment requires a DeepLearning.AI PRO membership.
Prerequisites
- Basic Python and comfort running Jupyter notebook cells (lessons are code-along)
- Familiarity with the general concept of large language models and API calls
- No prior Llama-specific experience required; the course is explicitly beginner level
Instructor
Amit Sangani
Instructor · DeepLearning.AI
Pros & Cons
Pros
- Authoritative source: built and taught by Amit Sangani, Senior Director of Partner Engineering at Meta, so the Llama-specific guidance comes straight from the team behind the models
- Genuinely hands-on: 7 code-along Jupyter notebooks let you run real prompts against Llama 2/3, Code Llama, and Llama Guard rather than just watching slides
- Free during the DeepLearning.AI platform beta and very short (~1h53m), making it a low-commitment way to learn the Llama prompting workflow
- Covers responsible-AI tooling (Llama Guard) that many short prompt-engineering courses skip entirely
- Well received: 4.4/5 on Class Central and Coursera (54 reviews) and positive forum feedback such as 'I just finished the course. Very well done.'
Cons
- Stability risk: the course was placed under maintenance for an extended period in late 2024 because of API/model deprecation issues 'outside the control of DLAI,' with no ETA given to affected learners, so code examples and availability may break
- Very narrow and beginner-level: it teaches Llama-specific prompting only and does not go into fine-tuning, RAG, agents, or general prompt-engineering theory
- Catalog mismatch: this entry is labeled 'vision,' but the actual course covers Llama 2/3 text, code, and safety prompting with no vision/multimodal content
- The graded quiz and shareable accomplishment require a paid DeepLearning.AI PRO membership; the free tier is audit-style
Alternatives To Consider
Frequently Asked Questions
Is Prompt Engineering with Llama 2 & 3 free?
Yes — Prompt Engineering with Llama 2 & 3 is free to access. Free to audit during the DeepLearning.AI learning-platform beta. A DeepLearning.AI PRO membership is required for the graded assignment and the earnable accomplishment/certificate; the course itself does not grant a standalone certificate. Note: it has been intermittently unavailable due to maintenance, so verify it is currently live before relying on it.
Who is Prompt Engineering with Llama 2 & 3 for?
Developers, data scientists, and technical learners who already understand basic Python and the idea of calling an LLM API, and who specifically want to learn how to prompt and choose among Meta's open Llama 2 and Llama 3 models, use Code Llama as a pair-programming helper, and add Llama Guard safety checks. DeepLearning.AI's own 'Who should join' note frames it broadly as anyone interested in prompt engineering who wants to try the Llama models.
What will you learn in Prompt Engineering with Llama 2 & 3?
How to call Llama 2 and Llama 3 chat models with a simple API and compare outputs across model sizes/versions; How to structure multi-turn conversations and format prompts so the model behaves predictably; Advanced prompting techniques: few-shot prompting (e.g., classifying message sentiment) and chain-of-thought prompting for logic problems; Using Code Llama as a pair-programming partner to write and improve code.
What are the prerequisites for Prompt Engineering with Llama 2 & 3?
Basic Python and comfort running Jupyter notebook cells (lessons are code-along); Familiarity with the general concept of large language models and API calls; No prior Llama-specific experience required; the course is explicitly beginner level.
Is Prompt Engineering with Llama 2 & 3 worth it?
It is a solid, free, official one-afternoon primer for people who already know basic Python/LLMs and want Llama-specific prompting habits. But it is short, beginner-only, tied to a specific (now partly deprecated) hosted API, and offers no broad prompt-engineering theory, so it is worth taking only if the Llama family is your target stack and you can confirm the course is currently live.
How we reviewed this course
This is an independent editorial assessment by Cursarium, based on DeepLearning.AI's published course materials and aggregated public learner feedback (last reviewed 2026-06). We have not independently completed the course. Links to providers are standard references, not paid placements.
Sources
- DeepLearning.AI - official course page (outline, instructor, 10 lessons, 1h53m, free-during-beta)
- Class Central - rating (4.4/5, 36 ratings) and syllabus
- Coursera - project page (4.4, ~53 reviews, beginner, <2 hours)
- DeepLearning.AI community - learner feedback ('Very well done') and course launch thread
- DeepLearning.AI community - maintenance/deprecation thread (course offline, no ETA)
- GitHub (ksm26) - notebook list mirroring the course lessons (L2-L8)