Cost tracking support for OpenAI Fast mode
Hassieb PakzadTrack OpenAI Fast mode costs automatically, including GPT-5.6 long-context rates.
You can now track OpenAI Fast mode spend accurately in Langfuse without ingesting costs manually. For supported OpenAI models, Langfuse automatically selects Fast mode pricing when a generation's model parameters contain service_tier: "fast" or service_tier: "priority".
- See the Fast mode surcharge in each generation's cost breakdown.
- Keep aggregate cost dashboards accurate when you mix Standard and Fast mode traffic.
- Apply the correct short- or long-context rate for GPT-5.6 calls.
OpenAI Fast mode offers faster inference at a per-token premium. OpenAI accepts both fast and the previous priority value for service_tier; for GPT-5.6 and earlier models, the response reports priority for either request value. Langfuse therefore matches both values. For GPT-5.6 models, it combines that service-tier match with input usage to select from the full Standard/Fast and short/long-context pricing matrix.
Pricing tiers are now more flexible for custom model definitions, too. In addition to matching numeric thresholds across usage details, a tier can match one or more exact values from any top-level model-parameter or metadata key. For example, you can price by the service_tier model parameter or top-level metadata keys such as region and model_provider. Configure these conditions under Project Settings > Models.