Ranked in these QuestionsQuestion Ranking
Pro Suggestions are truly context-aware
TabNine recognizes code intelligently by applying machine learning techniques to your codebase. It can generalize from only two examples, suggest completions based on lines of code before it, and even suggest completions for comments.
Pro Works out of the box
TabNine literally requires no configuration at all — it works as advertised out of the box.
Pro Paid features available for free when working with Rust code
The developers made paid features always available for free when auto-completing Rust code, in acknowledgment of how TabNine is made possible by the Rust ecosystem.
Pro Single-plugin install with no other dependencies
All you need to use TabNine is install the vim plugin. It doesn't require any external dependencies, binaries, packages, language servers, etc.
Pro Works with any language
TabNine can provide suggestions for any language, thanks to the way TabNine's completion engine works.
Pro Extremely fast
TabNine can suggest completions extremely fast — they pop-up immediately after you type a letter with no noticeable lag, compared to some other completion plugins. It promises ~20ms response time when auto-completing anything, and it's built on Rust where speed and memory-efficiency are first-class citizens of the language and ecosystem.
Pro Compatible with other editors
Apart from vim, you can use TabNine with Sublime Text, VS Code, and Atom.
Pro License keys are one-time purchases
You can purchase a license key for $29 and receive free updates — there's no subscription models of any sort. The license key can also be used multiple times, on various other supported editors, since they're not locked to a single installation or machine.
The TabNine binary required by the vim plugin is proprietary, and subject to TabNine's EULA.
Con Free version is severely limited
The free version can only really provide suggestions for small projects not exceeding 200KB in file sizes. Once you hit the limit, it only provide suggestions partially. This is due to the way TabNine works — it builds an index of your project and gives suggestions based on that, and 200KB is the index limit for the free version.
Con "Deep Learning" model is RAM hungry
It uses a lot of memory for a text-editor feature, and that's on top of whatever text editor you're using. Open a few instances of VS Code + TabNine + Firefox and you can easily cause your computer to start heavily swapping. (This only applies to the local ML model, not if you trust the cloud ML model instead.)