Strategies

parallem supports three execution strategies:

• Sync: executes each request immediately in normal Python control flow.

• Batch: collects requests for provider-side Batch API execution for a 50% discount. ParaLLeM's caching lets you resume where you left off. Great for high-throughput use cases.

• Concurrent: similar to async, but is not true async because await and .final_answer do not truly yield control to other processes. It still allows requests to be parallelized and executed concurrently.

See also

• Quickstart
• Ask
• MessageState
• Persistence