DEPLOY · NINETYFIVE
The fastest autocomplete.
Trained on your own code.
NinetyFive is Numerata's inference layer: custom models trained on your codebase, served with 50ms median latency, deployed inside your own trust boundary.
CORE CAPABILITIES
50ms median latency. That's almost as fast as your next keystroke.
CUSTOM MODELS
Trained on your own codebase, not a generic public model, so suggestions actually match your conventions and internal tools.
PRIORITY SUGGESTIONS
Faster, higher-priority completions for teams that depend on autocomplete as part of their daily workflow.
LOCAL INFERENCE
Run inference entirely on your own machine, no internet required. Coming soon.
YOUR CODE STAYS YOURS
All code sent to NinetyFive is subject to a 14-day retention policy, and is never used to train any model except your own.
Binary files, anything covered by .gitignore, and files processed with git-crypt are excluded automatically. For teams that need more, secure compute and secure training run entirely within your own infrastructure, the same private-by-design posture behind every layer of Numerata's stack.
WORKS WHERE YOU ALREADY DO
VS CODE SUBLIME INTELLIJ NEOVIM EMACS
PRICING
Free to start. Academic licenses available on request.
ULTRA
$10
per month
- Custom models
- Priority suggestions
- Local inference