specification: API Commons Rate Limits
specificationVersion: '0.1'
schema: https://raw.githubusercontent.com/api-evangelist/interface-research/main/schema/api-commons.yml#/$defs/RateLimits
provider: Tabby
providerId: tabby-ml
created: '2026-07-11'
modified: '2026-07-11'
reconciled: false
tags:
- AI Coding Assistant
- Code Completion
- Open Source
- Rate Limiting
- Quotas
description: >-
  Tabby does not publish fixed numeric API rate limits. Because the server is
  self-hosted, throughput is bounded by your own hardware - primarily the GPU (or
  CPU) serving the completion and chat models - rather than by a vendor-imposed
  per-minute request cap. Completion latency and tokens-per-second depend on the
  model size and device you configure. On hosted Team and Enterprise plans,
  capacity is governed by seat count and the managed infrastructure rather than a
  documented request-rate limit.
notes: >-
  No per-account or per-endpoint numeric limits are documented as of the review
  date. Tune performance by choosing an appropriately sized model and device, and
  scale horizontally by running additional Tabby instances behind a load
  balancer. Completion and chat endpoints return 501 when the corresponding model
  is not configured.
sources:
- https://tabby.tabbyml.com/docs/welcome/
- https://www.tabbyml.com/pricing
- https://github.com/TabbyML/tabby
responseCodes:
  notImplemented: 501
limits:
- name: Completion / Chat Throughput
  scope: deployment
  metric: requests
  limit: hardware-bound
  notes: Constrained by your own GPU/CPU and the configured model, not by TabbyML.
- name: API Request Rate
  scope: account
  metric: requests
  limit: not published
  notes: No fixed numeric request-rate limit is documented for the server API.
- name: Doc Ingestion
  scope: deployment
  metric: documents
  limit: not published
  notes: Documents are indexed asynchronously; throughput depends on your indexing resources.
- name: Hosted Seats
  scope: account
  metric: users
  limit: per plan
  notes: Community up to 5 users; Team up to 50 users; Enterprise unlimited.
policies:
- name: Self-Hosted Scaling
  description: Scale by selecting a larger device/model or running multiple Tabby instances behind a load balancer.
- name: Model Not Configured
  description: Completion and chat endpoints return 501 Not Implemented when their model is not configured, rather than throttling.
- name: Backoff Strategy
  description: Clients should implement exponential backoff with jitter when the server is saturated on constrained hardware.
maintainers:
- FN: Kin Lane
  email: kin@apievangelist.com