All learning paths

AI learning path

Frontier models and model selection

Choose models with evidence instead of leaderboard screenshots and social-media vibes.

Best for
Builders choosing between Claude, GPT, Gemini, Llama, Mistral, Cohere, DeepSeek, Qwen, Grok, Perplexity, and hosted open models
Level
Intermediate
Time
6-12 hours

Choose this when

You are comparing providers, migrating models, routing tasks, or balancing cost, speed, context, and quality.

You should be able to

You can build a small model-selection matrix and test candidates against your own tasks.

Checkpoint

Move on when you can justify model choice by task evidence, not brand preference.

Do

Learning sequence

This is the route through the topic. Watch and open the material inside the step where it is used.

Search resources Search videos

Step 1

Define the job

Separate coding, extraction, chat, multimodal, long-context, search, and agent tasks.

  • Task type
  • Risk
  • Budget

Watch here

The last six months in LLMs in five minutes

Simon Willison

Use this for current model landscape orientation before comparing docs.

Open here

OpenAI model guide

Model docs · OpenAI · Beginner to advanced

You need to choose between GPT-5.5, GPT-5.5 Pro, GPT-5.4 mini, GPT-5.4 nano, reasoning levels, tool support, and cost-sensitive API paths.

Open resource

Step 2

Compare primary docs

Read official model docs for context, modalities, tool support, pricing, and deprecations.

  • Capabilities
  • Limits
  • Pricing

Open here

Claude choosing a model

Guide · Anthropic · Intermediate

You need Anthropic's current guidance on balancing capability, speed, and cost before changing Claude models.

Open resource

Step 3

Run your own eval

Test candidate models on real examples before routing or migrating production workflows.

  • Eval set
  • Routing
  • Migration

Watch here

Open Challenges for AI Engineering

Simon Willison

Good context for model tradeoffs, migrations, and evidence-based selection.

Open here

Gemini API models

Model docs · Google AI for Developers · Beginner to advanced

You need to compare current Gemini 3.5 model families, context windows, modalities, stable releases, and experimental model IDs.

Open resource

OpenRouter models guide

Models guide · OpenRouter · Beginner to advanced

You need to compare many model families through one catalog before testing prompts across providers.

Open resource

Practice task

Build a five-row comparison table for two or three models using your own prompts, not generic benchmarks.

Reference

All resources in this path

Search resources

Step 1

OpenAI model guide

Model docs · OpenAI · Beginner to advanced

You need to choose between GPT-5.5, GPT-5.5 Pro, GPT-5.4 mini, GPT-5.4 nano, reasoning levels, tool support, and cost-sensitive API paths.

Step 2

Claude choosing a model

Guide · Anthropic · Intermediate

You need Anthropic's current guidance on balancing capability, speed, and cost before changing Claude models.

Step 3

Gemini API models

Model docs · Google AI for Developers · Beginner to advanced

You need to compare current Gemini 3.5 model families, context windows, modalities, stable releases, and experimental model IDs.

Step 3

OpenRouter models guide

Models guide · OpenRouter · Beginner to advanced

You need to compare many model families through one catalog before testing prompts across providers.

Educators to follow

Simon Willison profile photo

Simon Willison

Beginner to advanced

Read the recent model-roundup posts, then try the llm command-line tool with two or three different providers.

View educator
Swyx profile photo

Swyx

Intermediate to advanced

Watch AI Engineer talks for production patterns and tool choices.

View educator
Josh Pigford profile photo

Josh Pigford

Beginner to intermediate

Read the public notes and examples before deciding whether the paid material matches your business.

View educator