Phonic-1 debuts as top speech-to-speech model for agentic calls, trained on Together AI

graceisford · x · 2026-10-31

Together AI announces Phonic-1 from Phonic, billed as the most advanced speech-to-speech model for agentic tasks, running highly challenging calls across six continents. Traditional S2S models have traditionally sacrificed intelligence; Phonic-1 aims to close that gap. It was trained and served on Together AI and ranked as the #1 speech-to-speech model.

Related event: 10-person startup Phonic launches Phonic-1, tops speech-to-speech index and raises $27.5M Series A(16 posts)→

Original post →

More from Multimodal

• Magnific launches Move: conversational agentic motion graphics with every layer editable — charis_ai · 2026-10-31
• Director runs 'vibe directing' with Claude Code: per-shot approvals, $0.82 clips — DavidmComfort · 2026-10-31
• Ostris trains MOSS Image with Primal VAE: 4K at 1080p compute cost — Crazy-Repeat-2006 · 2026-10-31
• Creative director's wordless AI short: 119 shots distilled from hundreds of generations — Superb-Meeting7108 · 2026-10-31
• Moiré Pattern Enigma prompt: overlapping lines create shifting AI art — LudovicCreator · 2026-10-31
• AI video of the week: Jim Cramer must save America — mark_k · 2026-10-31

Multimodal channel →