Reference Limits — ML Junction docs

Field or endpoint · Limit
model · 1 through 150 characters
messages · At least one; must include a user message
output.max_tokens · 1 through 131,072
tools · Up to 128
idempotency_key · Up to 200 characters
routing.route · Up to 200 characters
only_providers / ignored_providers · Up to 20 entries each
fallback max_total_attempts · 1 through 10
fallback max_retries_per_route · 0 through 3
sticky ttl_seconds · 60 through 86,400
reasoning.continuation_token · 20 through 131,072 characters at request validation; encrypted payload is also constrained by the server continuation byte limit
reasoning continuation soft TTL · Deployment setting; default 86,400 seconds, allowed 60 through 31,536,000 seconds
usage window_days · 1 through 365
status snapshot window_hours · 1 through 720
status history/error window_hours · 1 through 2,160
graph bucket · hour or day

Canonical URL: https://mljunction.com/docs/reference-limits