Comparison of all code assistants!

code_assistants_comparison.md

	GitHub Copilot	Amazon Q Developer	Gemini Code Assist (Google Cloud)
Functionality	👍👍👍	👍👍👍	👍👍👍
Chat	Present	Present	Present
Code completion	Present	Present	Present
Other value-add functions in IDE (VSCode) Note that chat can be used to supplement some of these functions, but not indicated as directly supported from the UI	• Code explanation • Code fix (suggestions) • Generate unit tests • Generate docs	• Code explanation • Code refactor • Code fix (suggestions) • Code optimise (suggestions) • Does security vulnerability and code quality scanning (500 project scans per month on Pro tier)	• Code explanation • Generate unit tests • Google Cloud interoperability
Usability	👍👍👍	👍👍	👍👍
Multiple chat window support	It's slightly challenging to open multiple tabs, but still doable	Can open in multiple tabs with no visible limit	Limited to 1 chat in IDE
Context window size	~8000 tokens	4000 characters (Pro)	1 million
Chat history	Most of previous queries and responses can be reinstated in a new chat window	❌	❌
IDE Availability	• Visual Studio Code • Visual Studio • Vim and Neovim • Various JetBrains IDEs - GoLand, Android Studio, and 16 other IDEs • Azure Data Studio See full list here	• Visual Studio Code • Visual Studio • Various JetBrains IDEs - Goland, Android Studio, and 17 more IDEs • AWS coding environments See full list here	• Visual Studio Code • Various JetBrains IDEs - GoLand, Rubymine, and 9 other IDEs • Google Cloud - Cloud Shell Editor and Cloud Workstations See full list here
Programming Language Support	JavaScript, Java, Ruby, Python, and 9 other programming languages	JavaScript, Java, Python, Go, and 11 other programming languages	JavaScript, Java, Ruby, Python, and 18 other programming languages
Commands and shortcuts	• / for chat commands • @ for asking questions related to environment (can be mixed with # ) • # for adding context of VSCode, like referencing one or multiple files, in the prompt	• / for chat commands • @ for referencing environment like workspace (beta feature)	No clear UX functions like Copilot. To reference a file, you need to manually reference the file path in your prompt (making it substantially harder to ask questions, and less reliable if there are typos)
Reliability	❔	👍👍👍	👍👍👍
Service Uptime (SLA)	GitHub's SLA does not have any mention for Copilot	>= 99.9%	>= 99.9%
Performance	👍👍👍	👍👍👍	👍👍
Pricing (USD) / seat	Copilot Individual (Single User) • $10 per month • $100 per year Copilot Business (Organisation) • $19 per mth Copilot Enterprise (more custom use cases) • $39 per mth	Amazon Q Developer Pro Tier $19 per user per month	"Gemini Code Assist" paid tier $228 per user per year
Free usage coverage	• One time 30-day free trial. • Free for verified students, teachers, and maintainers of popular open source projects.	"Amazon Q Developer Free Tier" is a free tier for any user. Note the differences in support for Pro vs Free tier here.	"Gemini Code Assist" is free for one user per billing account, until November 8th, 2024.
Service Perfomance (accuracy of solution and speed)	In general, all three clients are indistinguishable in terms of functionality and performance in responses (I don't find that I am bothered by their responses). There was a period when Gemini CodeAssist did not allow answers about competing products like AWS, but this constraint has been lifted since.	-	-
Sustainability	👍👍	👍👍	👍👍
Security & Privacy	On data processing by GitHub Copilot, GitHub mentions that it doesn't uses Copilot Business or Enterprise for data to train Copilot, though it doesn't mention about Copilot Individual (the less pricier subscription for individuals).	On data protection in Amazon Q Developer, Amazon Q stores your questions, its responses, and additional context, such as console metadata and code in your IDE, to generate responses to your questions and you should not put confidential or sensitive information to Amazon Q. To ensure that your content is not used to train models used for Amazon Q, use Amazon Q Pro instead of free. In the pro tier, IP indemnity is also provided for, and you're opted out of data collection.	On data governance for Gemini, prompts and responses are not use to train models. Additionally, Gemini's prompt and responses are subjected to Google's Acceptable Use Policy, where unacceptable content are blocked.