In an Oct. 29 post, @moinnadeem introduced Phonic-1 as Phonic’s speech-to-speech model for agentic tasks and said the company had raised a $27.5 million Series A. The post named @firstround as lead investor, with @Lux_Capital, @felicis and Definition participating.
In a separate assessment, Artificial Analysis ranked Phonic 1 (High) first of 22 models on its Speech to Speech Index. Its score of 87.1 edged Grok Voice 3.0’s 86.9. The index averages speech reasoning, agentic performance, arena preference and task success, so the narrow overall lead does not mean Phonic topped every measure.
That distinction shows up in the evaluator’s Speech Agent Arena, where people have blind live voice conversations and choose their preferred model. Phonic 1 (High) ranked eighth of 30 for preference and fourth for task success, completing 92.3% of tasks. It ranked first at 79.9% on Tau Voice, a separate benchmark that tests policy-following and multi-step tool use in longer customer-service tasks. Artificial Analysis also measured an average 0.61 seconds to first audio on its Big Bench Audio benchmark, the fastest result among the index’s top five models.
Together AI stated that Phonic-1 was trained and served on its platform. Artificial Analysis said it evaluated the High configuration through the Phonic API and listed its price at $0.07 per minute.
Comments