Published on June 23, 2026, the index combines three benchmarks to compare native Speech-to-Speech models on speech reasoning, conversational dynamics, and agentic performance.
On June 23, 2026, Artificial Analysis introduced an index of native Speech-to-Speech models. The evaluation combines Big Bench Audio, a subset of Full Duplex Bench, and τ-Voice, giving each benchmark equal weight. A model must have valid results on all three datasets to be included.
The benchmark is intended to help engineers compare speech reasoning, conversational dynamics, and agentic performance. The available description gives the scope and inclusion criterion but no specific scores or results. Consult the original Artificial Analysis publication to verify the data, methodology, and any updates before using the index in a technical decision.