Run evals 50% cheaper with workflow evals

Braintrust introduced workflow evals that use OpenAI and Anthropic batch APIs to reduce evaluation costs by 50% for workloads that do not require immediate results, such as large regression suites, model comparisons, and delayed scoring jobs. Available in the TypeScript/JavaScript and Python SDKs, the feature coordinates submission and collection through developer-provided callbacks, while supporting polling or webhooks to detect completed batches. New WorkflowBatchTask and WorkflowBatchScorer components group cases into configurable batches and map returned results to individual items. Because providers may take up to 24 hours to complete requests, multi-stage workflows that batch generation and LLM judging must wait for generated outputs before judging begins. Runs can be started, monitored, and resumed later, with optional persistent Redis storage allowing another process to continue work after a batch has been submitted.

Use this post, company, and trend context to find content marketing opportunities, perform competitive analysis, or address product feature gaps via the Plushcap MCP server or the Plushcap API.