ElevenLabs brings Eleven v4, the top-ranked voice model, to OpenRouter

ElevenLabs' voice models went live on OpenRouter on October 7: eleven listings added between 15:06 and 15:18 UTC, nine text-to-speech models and two speech-to-text models. They include Eleven v4 and Eleven v4 Turbo, which ElevenLabs launched on September 28 and which now hold the top two places on the Artificial Analysis Text to Speech arena.

THE PRICE

OpenRouter prices text-to-speech per character and transcription per second of audio: - Eleven v4: $40 per million characters - Eleven v4 Turbo: $20 per million characters - Eleven v3 and Multilingual v2: $40 per million characters - Eleven v3 Conversational, Flash v2 and v2.5, Turbo v2 and v2.5: $20 per million characters - Scribe v2 and Scribe v2 Medical (speech-to-text): about $0.11 per hour of audio

On the same catalog, Alibaba's Qwen-Audio-3.0-TTS-Plus is also $20 per million characters and Microsoft's MAI-Voice-2.1 is $22.

THE RANKING

On the Artificial Analysis Text to Speech arena, Eleven v4 Turbo leads at 1,334 Elo and Eleven v4 is second at 1,321, ahead of Qwen-Audio-3.1-TTS-Plus (1,292), Cartesia Sonic 3.6 (1,278) and Gemini 3.8 Flash TTS (1,275). Arena Elo measures listener preference on general prompts, not fitness for any one job.

WHAT WE DO NOT KNOW

Every ElevenLabs endpoint carries a 50% discount flag in OpenRouter's API that the listing does not explain, so the billed price may be lower for now, for an unknown period. We found no announcement of the listing from either company at the time of writing. Eleven v4's launch figures, including its latency and blind-test preference rates, are ElevenLabs' own claims and were not checked here.