Phonic 1 Tops Artificial Analysis Speech-to-Speech Index at 87.1

On October 30, Artificial Analysis released its Speech to Speech Index, with startup Phonic's Phonic 1 (High) debuting at the top of 22 models with a score of 87.1, narrowly edging out Grok Voice 3.0 to become the strongest native speech model in current comprehensive evaluations.

Confirmed

• Overall leaderboard: Phonic 1 (High) ranks first among 22 models with 87.1, slightly ahead of second-place Grok Voice 3.0 (in the 86-point range)
• Latency: Average time to first audio on the Big Bench Audio benchmark is 0.61 seconds, about 25% faster than Grok Voice 3.0 (0.81 seconds), making it the fastest among the top five
• Pricing: $4.20 per hour of input audio ($0.07/minute), lower than Grok Voice 3.0 ($4.80) and GPT-Live-1 (Sol, low), roughly 40% of GPT-Realtime's price
• Speech reasoning: 98.9% on Big Bench Audio, ranking third, ahead of Grok Voice 3.0 (98.5%) and Gemini 3.8 Live Extended
• Agentic capability: First place on the Tau Voice agentic benchmark (simulating customer service scenarios, testing policy adherence and multi-step tool calling) with 79.9%, leading the second-place GPT-L series model by 5.4 points
• Human blind tests: Elo of 1012 in the Speech Agent Arena, popularity rank 8/30, near Cartesia Managed Agents

Why it matters

Phonic 1 leads on multiple dimensions at once—overall score, latency, pricing, and agentic tasks—showing that competition in speech models has expanded from conversational quality alone to cost, response speed, and tool-calling ability. However, its 8th-place finish in the human blind-test Arena suggests automated benchmarks and user preferences are not fully aligned, and the evaluation system itself remains contentious.

2026-10-30 ~ 2026-10-30 · 8 related posts

• [source] Phonic 1 Tops Artificial Analysis Speech-to-Speech Index at 87.1, Edging Grok Voice 3.0 — ArtificialAnlys · 2026-10-30
• [source] Speech Agent Arena: Phonic 1 Ranks #8 in Preference with 92.3% Task Success — ArtificialAnlys · 2026-10-30
• [source] Phonic 1 Leads Tau Voice Agentic Benchmark at 79.9%, 5.4 Points Clear — ArtificialAnlys · 2026-10-30
• Phonic 1 Scores 98.9% on Big Bench Audio, Tops Tau Voice Agentic Benchmark — ArtificialAnlys · 2026-10-30
• Phonic 1 Hits 0.61s Time-to-First-Audio, Fastest of the Top 5 Speech Models — ArtificialAnlys · 2026-10-30
• Phonic 1 Costs $4.20/Hour of Input Audio, Undercutting Grok and Most GPT-Live Tiers — ArtificialAnlys · 2026-10-30

2 near-duplicate retellings: ArtificialAnlys · graceisford