LLM·Dex
Newsletter
Use case

Best Speech-to-Text Models in 2026

Audio transcription for meetings, podcasts, voice interfaces.

Updated

How we ranked

  • Word error rate (WER) on clean and noisy audio
  • Speaker diarization quality
  • Multilingual coverage
  • Streaming vs. batch latency
  • Cost per minute of audio

Read the full methodology for our sourcing and ranking standards.

Whisper and its many descendants set the open baseline; commercial leaders now include Deepgram, AssemblyAI, and provider-bundled offerings. Recent multimodal LLMs (Gemini, GPT-5) accept native audio input but still trail dedicated STT on WER for specialized accents.

For real-time voice agents, latency matters more than absolute WER. Deepgram Nova and Cartesia lead the streaming-WER tradeoff.

For batch transcription, OpenAI Whisper Large v3 is hard to beat on cost-per-minute.

This category is currently tracked but has no LLM-tier ranking, the leaders live in the dedicated tools section. See our alternatives index for the full picks.

How to choose

Don't pick on the headline ranking alone. Run your top two picks on a representative sample of your own workload and compare. The numbers in this list are sound, but task-specific quality varies in ways no benchmark fully captures. The criteria above are the right axes to evaluate on, but the weighting depends on your stack.

  • Cost-sensitive workloads, start with the cheapest of the top three; escalate only if quality is the bottleneck.
  • Privacy-sensitive workloads, filter to open-weight picks above. They're labeled with a green badge.
  • Latency-sensitive workloads, see the Fastest LLMs list, which can override task-specific picks.

Frequently asked

  • How are these rankings determined?
    We rank by the criteria listed at the top of this page: Word error rate (WER) on clean and noisy audio; Speaker diarization quality; Multilingual coverage. Where two models are close, we prefer the one with stronger production deployment evidence at the time of writing. Read the full methodology for our standards.
  • Are open-source models on this list?
    Yes where they're competitive. Each entry below shows whether the model ships open weights and under what license.
  • How often is this list updated?
    Weekly. New launches that affect the ranking get reflected within seven days. The "last updated" stamp at the top of the page reflects the most recent dataset commit.

Related guides

Friday digest

The week's AI launches, in your inbox.

One short email every Friday, new models, leaks, and quietly-shipped APIs you missed.