Speech Agent Arena: Phonic 1 Ranks #8 in Preference with 92.3% Task Success
ArtificialAnlys · x · 2026-10-30
Supplementary data from Artificial Analysis: in the Speech Agent Arena, where humans hold blind live voice conversations and pick their preferred model, Phonic 1 (High) debuts at #8 of 30 with an Elo of 1012 — behind Cartesia Managed Agents, Gemini 3.1 Flash Live Minimal and Gemini 3.8 Live, ahead of Eleven Realtime v1 Max. It completes 92.3% of tasks successfully, #4 behind Grok Voice 3.0, Grok Voice 2.0 and Gemini 3.8 Live. Full Arena Elo and task-success charts included.
Related event: Phonic 1 Tops Artificial Analysis Speech-to-Speech Index at 87.1(8 posts)→
More from Models
- Barclays: Healthcare firm fine-tuned open model to 93% accuracy in 3 days, 98% cheaper than frontier — eliano · 2026-10-30
- Qwen3.5-9B appears to have a 'sunk cost neuron' that derails every answer — adonis_singh · 2026-10-30
- TEM: Transformer Energy-Based Models Claim to Outscale Diffusion Without Guidance — du_yilun · 2026-10-30
- ChatGPT still won't let you change your account's phone number — functionaldude · 2026-10-30
- 4,700 blind judgments show DeepSeek/GLM with a custom design harness match Claude Code + Sonnet 5.5 — JannVanDam · 2026-10-30
- Eval lead on benchmaxxing concerns: "we will always have newer tasks" — cHHillee · 2026-10-30