Roadmap

The next Universal-3 Pro release. Native-language coverage grows from 6 to 19, with accuracy gains on the six core languages.

A new model option priced the same as u3-pro, not an automatic drop-in. The current model stays available as a pinned snapshot.

New languages: Japanese, Vietnamese, Arabic, Dutch, Swedish, Hindi, Norwegian, Finnish, Danish, Urdu, Hebrew, Mandarin, Turkish.

Universal-3 Pro transcription roughly twice as fast end-to-end. Phase one is live, already cutting turnaround time 30-80%.

Much better speaker labeling in noisy, multi-speaker audio. Targets the two most-reported errors: mislabeled short replies like “yeah” and “uh-huh”, and speaker turns that don’t line up with punctuation.

A new model option priced the same as u3-pro. The current model stays available as a pinned snapshot.

Open-source speech-to-text models served directly through our API, for the languages and domains they specialize in.

Recognize the same speaker across recordings, not just within a single file. For meetings, call centers, and cross-session analytics.

Synchronous HTTP transcription on u3-pro, ~134ms p50 latency, for voice-agent calls that need a transcript in a single request.

On-premise universal-3-pro for regulated environments with strict data-residency requirements.

The next major accuracy and capability release after Universal-3.5 Pro.

Our new flagship realtime model gives every turn the context a conversation actually has. The model takes direction from your agent, keeps a rolling memory of the call on its own, and hears the speaker instead of the room, so spelled-out emails, account IDs, and one-word confirmations come out right. It runs in 19 languages at flagship accuracy with mid-sentence code-switching, plus a new language_code parameter to commit to one language when you already know it.

A new model option priced the same as u3-pro, not an automatic drop-in. The current model stays available as a pinned snapshot.

New languages: Japanese, Vietnamese, Arabic, Dutch, Swedish, Hindi, Norwegian, Finnish, Danish, Urdu, Hebrew, Mandarin, Turkish.

A fast, cost-efficient realtime model for notetaking and meeting intelligence. Tuned for long-form audio where throughput and stable accuracy over multi-hour sessions matter more than latency.

On-premise universal-realtime-3-pro for regulated environments with strict data-residency requirements.

The next-generation realtime model for voice agents. Targets the lowest turn latency and the strongest handling of voice-agent audio (noise, interruptions, hesitation, accented speech), with instruction-following strong enough to replace today’s STT + LLM + TTS stack. Multilingual across 15+ native languages and the foundation for our speech-to-speech architecture.

A single native model that replaces today’s Voice Agent pipeline (STT, LLM, TTS) with a unified Realtime Speech LLM. Tighter latency, better prosody, and more natural interruption handling than orchestrated stacks.

Direct Twilio SIP and voice connectivity, without customer-side LiveKit plumbing.

Full programmatic control of voice agents at the edge. Create, version, and manage agents as code, persist and retrieve sessions (events, transcripts, tool calls), and run webhooks and tool calls at the edge instead of round-tripping to origin.

Official client libraries, starting with Python and TypeScript.

A custom turn-detection model trained on Universal 3 Pro streaming output. Reduces false endpointing and handles pauses, hesitations, and overlapping speech better.

A standalone text-to-speech model for production voice workloads. Low time-to-first-byte, voice prompting, and accurate delivery of phone numbers, emails, and named entities that today’s TTS struggles with.

Open-source text-to-speech models served through our API alongside Universal TTS, for the languages, styles, and domains they specialize in.

Summarization via the LLM Gateway, replacing legacy LeMUR summaries, using frontier models with automatic fallbacks.

Sharper chapter boundaries and titles via the LLM Gateway, with better topic segmentation on long-form content.

Accuracy improvements to Speaker ID, Translation, and Custom Formatting. Translation covers both streaming and pre-recorded audio, for workflows where the spoken language differs from the output.

Static find-and-replace redaction (redact_pii_static_entities) for known terms, with LLM-powered category redaction following in Q3/Q4.

Automatic LLM-based transcript correction with no user prompts, generalizing the Medical Mode pattern to any domain.

Better keyterm prompting in every supported language via LLM Gateway post-processing, closing the gap with English for Spanish, German, French, Portuguese, and Italian.

Detect speaker emotions and shifts in input audio, distinct from TTS-side Emotion and Style Tagging. For therapy, CX scoring, and compliance monitoring.

Ongoing catalog expansion. The Gateway supports 24 models across Anthropic, OpenAI, Google, Qwen, and Kimi, with DeepSeek, Mistral, Llama, and Cohere next.

Priority, standard, and flex request tiers for per-request cost and latency control.

A public version of our internal evaluation dashboard, covering 30+ competitors and the community models we serve across dozens of metrics. Verify accuracy and pick the right model per language and domain.

Open-source release of the dashboard and tooling, so anyone can reproduce our benchmarks on their own data.

A realistic Voice-AI evaluation set of telephony, meeting, and voice-agent audio with ground-truth transcripts, used to grade every model on the leaderboard. Not a thirty-second YouTube clip or LibriSpeech.

SAML and OIDC, as a fast-follow to multi-user accounts.

Guided setup for new accounts: model selection, API key creation, a first-request walkthrough, and best-practice defaults.

Hard spend caps and team budgets, beyond today’s soft alerts.

Deeper dashboard observability: P50 and P95 turnaround time, webhook delivery stats, uptime, and latency histograms.

Programmatic access to billing and usage data.

Configurable alert thresholds and cadences, including daily alerts and custom triggers.

Card support for EU PSD2 Strong Customer Authentication.