The router for voice AI. 🌊 Every speech, LLM and voice model, benchmarked by language, behind one API.
Joined July 2026
- Tweets141
- Following15
- Followers405
- Likes125
@omihealth omi-medical-1 is live on Speko
Median time to the first transcribed word is 962 ms, second fastest of 37 models we measure, with 4.0% word error rate on top
Built for medical English, now one API call away
speko.ai/benchmarks
Speko is the #1 rated MCP server on Manufact's leaderboard
97/100, the top score out of 1,456 verified servers
Connect it to ChatGPT, Claude or Grok and your assistant can make phone calls and build voice agents
manufact.com/leaderboard/spe…
Speko Week of Voice (Day 5/5)
Get 185 hours of voice AI for your startup
We’re giving selected startups $1,000 in Speko voice credits through our Startup Program
Build with models from 30 providers, including ElevenLabs, OpenAI, Anthropic, Deepgram and Google, through one API
Batch 01 teams also get a place in our startup showcase and features on our X and LinkedIn as they ship
An idea is enough to apply. Tell us what you want to build: speko.ai/startups?ref=x
3
182
@FishAudio transcribe-stream-1 posted the lowest error rate on real conversations of any model on Speko Benchmarks
11.5% word error rate across 60 speakers talking naturally, ahead of Qwen3-ASR, Inworld, ElevenLabs and OpenAI
The top four are close, so it's a race worth watching
Speko Week of Voice (Day 4/5)
Speko is now in ChatGPT, Claude and Grok Bot.
Your assistant can make phone calls, create voice agents and test them - all from your chat.
Ask it to call a dentist and check availability. Or set up a receptionist that answers calls and books appointments for yours.
Connect Speko to the assistant you already use: speko.ai/integrations
Speko Week of Voice (Day 3/5)
Introducing Speko Voice Agents. Describe the voice agent you need and create it in under 10 seconds.
Speko generates its instructions and recommends speech recognition, language and text-to-speech models using benchmark results for your language.
Connect your calendar, CRM and custom tools so your agent can book appointments and update customer records during calls.
You can talk to the agent directly in your browser, customize its voice and behavior, and connect it to your phone system.
We measured Soniox v6 on our speech-to-text board: 8.6% word error rate on conversational speech, down from 9.5% on v5.
It improves on read speech (5.4%) and accented conversation (14.7%) too.
Full results: benchmarks.speko.ai
5
12
374
Speko Week of Voice (Day 2/5)
100+ voice models benchmarked.
Introducing Speko Benchmarks, your source of truth for voice AI models.
Compare speech-to-text, text-to-speech, LLMs and speech-to-speech on accuracy, latency, quality and cost. Explore multilingual performance, code-switching and turn-taking.
New voice model? Check Speko.
It's Week of Voice at Speko (Day 1/5)
Five launches in five days, one every morning around 9:00 AM PT.
Day 1 is out now: pitch your startup to our AI VC 👇
Day 1: Pitch the VC
Call +1 (602) 847-9960 and pitch Morgan, our AI VC, in 60 seconds. You get a couple of hard questions and a score out of 10.
Leaderboard: pitch.speko.ai
1
8
369
We built Polymarket for Voice AI.
Will ElevenLabs v5 drop in the next two months? Will Deepgram Nova-4? Will any transcriber finally get under 12 errors per 100 words?
Place your predictions on markets.speko.ai
Vietnamese is now on the Speko STT board.
Soniox ranks 21st of 35 for live English. On live Vietnamese it is 2nd of 17, with a 2.3% word error.
In contrast, the model ranked 2nd for live English drops to 14th.
The best model depends on the language your callers speak.
Full results: benchmarks.speko.ai/stt-mult…
2
136
Congrats to @GradiumAI on open-sourcing Phonon. Phonon-31m is out under Apache 2.0, small enough to run on device.
A new frontier for on-device speech.
We're open-sourcing Phonon, a suite of TTS models under 50m parameters, setting a new standard for naturalness and accuracy across languages, with the broadest inference support to date.
Phonon-31m is available under Apache 2.0.
gradium.ai/phonon🧵
Grok Voice 3.0 by @xai is live on Speko.
It answers the caller in 1.6s, the fastest of any S2S model we time, and has the top task capability on our board: 0.94.
Try it: platform.speko.ai
Speko Translate is now live!
translate.speko.ai
Introducing Speko Translate.
Speko Translate calls any phone number from your browser and interprets the call live in both directions, in 51 languages, using the Speko calling API and the Speko Router.
Never put off a phone call because of the language again.
Try it now at translate.speko.ai
15
535
The @speko_ai MCP server is #1 in Developer tools on the Manufact MCP catalog.
99/100, grade A+.
Connect it to Claude, ChatGPT, or Cursor and build voice agents from your editor: speko.ai/developers
The fastest Thai voice in Asia.
Paxa Labs starts speaking in 89ms from Singapore. The next fastest takes 177ms. ElevenLabs takes 789ms.
It's 2nd of 14 on getting the tones right, too.
On Speko now. benchmarks.speko.ai
You can now bring your Twilio numbers to Speko.
Paste your Account SID and API key. Speko builds the SIP trunk in your Twilio account and attaches your numbers. No forwarding, no second console.
Try it: platform.speko.ai
11
499
Palabra TTS is live on Speko, at 50% off.
Realtime speech synthesis built for live translation and dubbing, now routed through one API alongside 15 other text-to-speech models.
Try it: speko.ai/models
ElevenLabs v4 and v4 Turbo just posted the highest robustness we have ever measured on our TTS board
v4 also leads on naturalness.
Turbo trades some naturalness for 205ms latency at half the price.
See the board: benchmarks.speko.ai/tts
Both are now routable on Speko. Pick eleven_v4 or eleven_v4_turbo on platform.speko.ai, same as every other model on the gateway.
1
131