Conformer-2

Conformer-2 ASR trained on 1.1M hours of English audio

>93.3% accuracy with industry-low WER and fewer hallucinations

Real-time streaming transcription with low latency

Speaker diarization, entity detection, and sentiment analysis

Universal model with 99 languages, auto-detect, and code-switching

LLM Gateway: single API for 15+ LLMs with unified billing

Voice AI Guardrails: PII redaction and profanity filtering

Usage-based pricing with no long-term contracts

Enterprise security: SOC 2, ISO 27001, HIPAA BAA, PCI-DSS

Flexible deployment: cloud, VPC, on‑prem, AWS Marketplace

Scales to 600M+ inference and 840M+ API calls per month

Processes 40TB+ of audio daily with uncapped concurrency

Up to 53.7–55% faster inference latency vs. Conformer-1

31.7% better alphanumeric recognition; lower CER variance

6.8% improvement on proper nouns for downstream GenAI quality

12%+ robustness gains to noise; 43% fewer errors on noisy sets

Accurate timestamps (≈200 ms) with improved ITN and truecasing

Easy developer integration and high-touch enterprise support