Model Ecosystem Map

Need help?

Ask in Discord Open a GitHub issue Support guide

Summary

The model ecosystem is easier to understand when it is grouped by what a learner can do with each family: chat and reasoning, multimodal input, generation, open-weight deployment, or regional cloud access. This page is a high-level map for choosing a starting point, not a benchmark ranking.

Why It Matters

Students and early builders often hear model names before they understand the product shape behind them. A useful map should answer four questions:

What kind of work can this model family support?
How do I access it: consumer app, API, open weights, or cloud platform?
Is it better for learning, prototyping, or deployment?
What should I check before using it in a real project?

Architecture Diagram

Provider Map

Provider family	What students should remember	Common access pattern	Good first use
OpenAI	Broad model platform across text, reasoning, image, audio, video, realtime, and agentic development surfaces.	ChatGPT for app use; OpenAI API for builders.	General assistants, coding help, multimodal prototypes, and agent workflows.
Anthropic Claude	Strong document, reasoning, coding, and workspace-oriented assistant family with API and app surfaces.	Claude app and Claude API.	Long-document review, careful writing, coding assistance, and artifact-style outputs.
Google Gemini	Google model family with API access and strong integration with Google AI Studio and Google Cloud surfaces.	Gemini app, Google AI Studio, Gemini API, and cloud deployment paths.	Multimodal experiments, search-grounded prototypes, and Google-stack applications.
Meta Llama	Open-weight model ecosystem that can be accessed through Meta and partner channels.	Download, partner hosting, or cloud/provider APIs.	Learning open-weight tradeoffs, local experimentation, and portability-minded builds.
Alibaba Cloud Qwen / Model Studio	China-linked model platform with Qwen, multimodal options, coding models, and OpenAI-compatible API patterns.	Alibaba Cloud Model Studio and DashScope-style API access.	China-aware app prototypes, Qwen experiments, and cloud-hosted model access.
Baidu Qianfan / ERNIE	China cloud model platform covering ERNIE, DeepSeek, Qwen-linked options, multimodal generation, search, and app builder surfaces.	Baidu Qianfan ModelBuilder and AppBuilder.	Chinese-language product experiments, enterprise app building, and multimodal exploration.
Moonshot Kimi	Kimi API family with long-context text and vision-oriented model options.	Kimi app and Moonshot API.	Chinese-language long-context work, document review, and early API experiments.

Capability Map

Capability	What it means	Typical model families to inspect
Chat and reasoning	Answers questions, drafts text, explains concepts, plans, and solves multi-step tasks.	OpenAI, Claude, Gemini, Kimi, Qianfan, Model Studio.
Vision and file understanding	Reads images, screenshots, diagrams, PDFs, and other uploaded materials.	OpenAI, Claude, Gemini, Qianfan, Model Studio, Kimi vision options.
Image, audio, and video generation	Creates or transforms media rather than only reading it.	OpenAI specialized models, Gemini ecosystem tools, Qianfan, Model Studio.
Tool use and agent workflows	Calls functions, uses tools, or connects to external systems.	OpenAI API, Claude API, Gemini API, Model Studio, Qianfan.
Open-weight deployment	Lets teams study, host, tune, or run models outside a single hosted app.	Llama, Qwen open-source editions, and partner-hosted open model catalogs.
Regional platform access	Helps teams match language, compliance, data residency, billing, and local ecosystem needs.	Alibaba Cloud Model Studio, Baidu Qianfan, Moonshot Kimi, cloud partners.

Choosing A Starting Point

Use the simplest surface that teaches the right lesson.

For first exposure, start with a consumer assistant such as ChatGPT, Claude, Gemini, or Kimi and focus on task design.
For API learning, start with one model provider and build a small request-response app before comparing vendors.
For multimodal learning, test one input type at a time: image, document, audio, or video.
For open-weight learning, start with why portability, local control, or licensing matters before selecting a model.
For China-linked deployment, inspect Qianfan, Model Studio, and Kimi from the beginning instead of treating them as late substitutes.

Common Mistakes

Treating the newest model name as automatically better for every task.
Comparing app features with API features as if they are the same product.
Ignoring pricing, rate limits, region availability, safety policy, and data controls until after a prototype works.
Choosing an open-weight model for portability without budgeting for hosting, evaluation, monitoring, and updates.
Choosing a regional platform only for language coverage instead of checking deployment, billing, and support requirements.

Suggested Class Exercise

Pick one task, such as “summarize a course reading and produce a study quiz.” Ask students to compare three access patterns:

a consumer assistant
a hosted API
an open-weight or regional cloud option

The output should be a short table: task quality, ease of setup, cost or usage limits, and what a production team would need to verify next.

Foundations

Patterns

Systems

Ecosystem

Case studies

Radar

Summary

Why It Matters

Architecture Diagram

Provider Map

Capability Map

Choosing A Starting Point

Common Mistakes

Suggested Class Exercise

Citations

Reading Extensions

Update Log

​Summary

​Why It Matters

​Architecture Diagram

​Provider Map

​Capability Map

​Choosing A Starting Point

​Common Mistakes

​Suggested Class Exercise

​Citations

​Reading Extensions

​Update Log

Summary

Why It Matters

Architecture Diagram

Provider Map

Capability Map

Choosing A Starting Point

Common Mistakes

Suggested Class Exercise

Citations

Reading Extensions

Update Log