LoginSign Up
← Back to all sparks
G

Gladia

AI-ASSISTANTS
Velocity3.8

Speech-to-text and audio intelligence API for transcription and analysis

www.gladia.io ↗

Gladia anchors on a new flagship STT model while stacking compliance and developer tooling.

speech-to-textmodelsbenchmarkscompliancemultilingualsdk
◆Current state
Gladia is a speech-to-text API vendor, and its recent cadence centers on model accuracy and trust. Solaria-3 is the new flagship, tuned for noisy, conversational production audio with stronger entity recognition; it follows measurable accuracy work like a 3x Hebrew improvement and an open, reproducible benchmark. Around the model, Gladia has shipped an async SDK, a multilingual normalization library, and refreshed SOC 2, HIPAA, and ISO certifications.
◆Where it's heading
Two tracks run in parallel: pushing recognition accuracy on real-world audio, and building the enterprise trust surface (certifications, open benchmarks) that wins regulated buyers. The Audio-to-LLM path hints at moving up the stack from transcription toward audio intelligence.
◆Prediction
Expect Solaria to keep iterating on accuracy and language coverage, with continued emphasis on transparent benchmarks as a differentiator against larger STT providers.

◆Recent moves

  1. 19d ago

    Solaria-3: Our new speech-to-text model

    ⚡ SPARK

    Solaria-3 is the new flagship speech-to-text model, tuned for noisy, fast, conversational audio with higher precision on names and business entities — the centerpiece of Gladia's accuracy-first strategy.

    View source ↗
  2. 1mo ago

    SOC 2 Type II & HIPAA Renewal

    SOC 2 Type II and HIPAA certifications were renewed, maintaining the existing compliance posture. Important for enterprise procurement but not a product capability change.

    View source ↗
  3. 1mo ago

    AI Meeting Assistant Market Map

    A published market map of the meeting-assistant space is marketing and category commentary, not a product change. It signals where Gladia sees its customers operating but ships nothing to users.

    View source ↗
  4. 2mo ago

    Multilingual Normalization Library

    The open-source normalization library adds French, German, Spanish, Italian, and Dutch with language-specific number expansion — concrete multilingual depth that complements the accuracy work on the core models.

    View source ↗
  5. 2mo ago

    Asynchronous SDK

    An official 1.0.0 SDK for the async STT API in TypeScript and Python cuts integration boilerplate, lowering the bar for developers adopting Gladia's transcription.

    View source ↗
  6. 2mo ago

    Audio to LLM is now generally available

    Audio-to-LLM reaches general availability, giving one API path from audio to transcript to insight. It points up the stack toward audio intelligence, though it's the GA of an existing capability rather than a new direction.

    View source ↗