promptfoo 0.123.0 adds native audio grading and new frontier-model provider support

promptfoo 0.123.0 (Sept 10, 2026) ships native-audio grading via llm-rubric, adds Claude Fable/Mythos 5.1 and GPT-6 provider support, and defaults GPT-5.6+ to the Responses API (breaking change) in this widely-used LLM eval/red-team OSS tool.

Routine but continuous feature cadence for a leading open-source LLM security-testing/eval framework, keeping pace with new frontier model releases used in enterprise red-teaming.

Teams already using promptfoo for LLM eval/red-teaming should review the breaking Responses-API default change before upgrading.