Conversational voice and audio understanding / Benchmark leaderboard

τ-Voice / τ³ full-duplex voice

Spoken service conversations with tools, interruptions and simultaneous audio.

No published company result yet.

This benchmark is in the research directory. Its task package, adapter and grading protocol need qualification before a hosted run can be offered.

Arrange an evaluation for your company →

What this benchmark measures

Metrics

Task reward plus separately logged latency, turn-taking and interruption recovery.

Execution requirements

Native real-time audio adapter and benchmark tool/user simulation.

Scope and limitations

This measures the complete voice agent, not TTS alone. The upstream supported-provider list is narrower than the prospect list; adapters are needed.

Evaluation availability

Catalog entry. Request a managed evaluation to qualify your agent interface and the benchmark’s native grading requirements.

Sources and company fit

Companies whose products may fit

Research recommendations based on product capabilities. These companies have not necessarily run this benchmark or integrated with Blobfish.

Compatibility notes for each company

Bland AI: Full voice-stack adapter

Cartesia: TTS API; full cloning protocol conditional

Cresta: Full voice-stack adapter

Decagon: Capability-aligned; adapter/access to qualify

Deepgram: Capability-aligned; adapter/access to qualify

ElevenLabs: TTS API; full cloning protocol conditional

Hume AI: Capability-aligned; adapter/access to qualify

Parloa: Full voice-stack adapter

PolyAI: Full voice-stack adapter

Retell AI: Full voice-stack adapter

Sierra: Capability-aligned; adapter/access to qualify

Vapi: Full voice-stack adapter