diff --git a/skills/together-chat-completions/SKILL.md b/skills/together-chat-completions/SKILL.md
index a636ab5..0d117fb 100644
--- a/skills/together-chat-completions/SKILL.md
+++ b/skills/together-chat-completions/SKILL.md
@@ -35,6 +35,7 @@ clearly offline batch processing, vector retrieval, model training, or infrastru
 - Use `together-fine-tuning` when the user wants to train or adapt a model
 - Use `together-dedicated-endpoints` when the user needs always-on single-tenant hosting
 - Use `together-dedicated-containers` or `together-gpu-clusters` for custom infrastructure
+- For production stock-model workloads that need a defined SLA (committed throughput and reliability) without managing hardware, point users to Together's [provisioned throughput](https://docs.together.ai/docs/inference/provisioned-throughput) tier (reserved PTU capacity, one-month minimum term, contact sales; uses the same chat/completions API surface)
 
 ## Quick Routing
 
diff --git a/skills/together-chat-completions/references/api-parameters.md b/skills/together-chat-completions/references/api-parameters.md
index ff2e06c..ccf3e96 100644
--- a/skills/together-chat-completions/references/api-parameters.md
+++ b/skills/together-chat-completions/references/api-parameters.md
@@ -389,7 +389,7 @@ Best practices:
 - plan against the latest headers instead of a hard-coded RPM table
 - keep traffic steady instead of bursty
 - use batch inference for high-volume offline jobs
-- use dedicated endpoints for strict capacity or SLA requirements
+- for strict capacity or SLA requirements, use provisioned throughput (reserved capacity on supported stock models with a defined throughput and reliability SLA, contact sales) or dedicated endpoints (single-tenant GPUs, self-serve)
 
 ## Debug Mode