VoiceBench — Results

Rank	Model	Score
1	gemini-3-1-pro	94.3
2	inkling	91.4
3	lfm3-omni-2b	64.73

VoiceBench — Phases

1 phaseActive

Phase 1·

VoiceBench

Jul 16, 2026

Current

About

TACL'26 benchmark for LLM-based voice assistants (Chen et al., 2024). Tests spoken instruction following, QA, and safety with real and synthetic audio across diverse speakers and e…

Pages

Phases: 1Active
Participants: 1
Submissions: 3
Last Modified: Jul 16, 2026