Sampling & Reasoning — ML Junction docs

Field · Accepted values · Notes
sampling.temperature · 0 through 2 · Higher values increase randomness; route support varies
sampling.top_p · Greater than 0 through 1 · Nucleus sampling; avoid changing both temperature and top_p without evaluation
sampling.seed · Integer · Requests determinism where supported; not a universal guarantee
reasoning.enabled · true or false · Requests a reasoning-capable route when paired with a requirement
reasoning.effort · none, minimal, low, medium, high, xhigh, max, auto · Portable coarse alias; intensity is more precise
reasoning.intensity · 0-100 · Mapped through the selected model route’s DB-backed native levels
reasoning.summary · none, auto, concise, detailed · Requests a summary where the provider supports one
reasoning.context · auto, current_turn, all_turns · Selects the requested reasoning context scope where supported
reasoning.continuation_token · Opaque gwrt_v3 string · Optional route-affinity ticket for provider-native continuation
reasoning.input_type · user_message, tool_results, dynamic_context_update · Describes what the current continuation input represents
reasoning.pending_turn_action · continue, cancel, fork · Resolves a pending tool loop

Supported parameters are route-specific. GET /v1/models/{model}/routes returns capabilities.supported_params for the exact provider route. With routing.strict_params set to true, unsupported requested parameters disqualify a route instead of being silently dropped.

Reasoning models may intentionally ignore temperature or expose a different token-control parameter. Use the capability list and strict_params for deterministic production behavior. See Reasoning Continuation for complete history replay, expiration, provider outages, and tool loops.

Canonical URL: https://mljunction.com/docs/sampling-reasoning