--- title: "The Router for Voice AI" description: "Hosted Router for managed STT, LLM and TTS routing, plus an open customer-side Gateway for provider-direct voice workloads." canonical: "https://speko.ai/" last-updated: "2026-08-24" --- > ## Speko page index > The complete index of every page on this site is at: https://speko.ai/llms.txt > Read it before exploring further. It lists the exact Markdown URL for every canonical HTML page. # The Router for Voice AI [Backed by Y Combinator](https://www.ycombinator.com/companies/speko) Every speech model, benchmarked language by language, wired into one API. [Get API key](https://platform.speko.ai/sign-in) ## Voice agent platform Build, test and run voice agents, with phone numbers, messaging and monitoring in one console. Press the sphere to call [Build your own](https://platform.speko.ai/sign-in) Maple Street Dental 1. Front desk Hi, thanks for calling Maple Street Dental. How can I help? 2. You Can I come in Saturday morning? ![The Speko console overview: sessions this week, usage and the busiest agents](https://speko.ai/_next/image?url=%2Fproduct%2Fconsole-overview-light.webp&w=3840&q=75&dpl=dpl_3VYqGmx5y2Z74qhNG8NNUX9s69pp) ![A call in the Speko console: transcript with per-turn latency, a transfer, and the recording](https://speko.ai/_next/image?url=%2Fproduct%2Fconsole-transcript-light.webp&w=3840&q=75&dpl=dpl_3VYqGmx5y2Z74qhNG8NNUX9s69pp) ### Build, test and run voice agents in one platform [Start building](https://platform.speko.ai/sign-in) Design the call, give it a phone number, test it against simulated callers and score every live call. Front desk · Workflow Draft v13 ### Workflows Map the call step by step. Calendars, CRMs, your APIs and docs plug into any step. Buy a number Area code 415 +1 (415) 555‑0134 Buy +1 (415) 555‑0172 Buy +1 (415) 555‑0190 Buy Your numbers +1 (628) 555‑0118 Front desk Inbound +1 (628) 555‑0145 Recalls Call list · 212 of 480 Twilio SIP trunk Front desk Your own number ### Phone numbers Buy a number in a click or bring yours over SIP. Inbound, outbound and call lists. Evaluation · Draft v13 19 of 20 passed Reschedules twice, then cancels Impatient caller Pass Books a cleaning in Spanish Second language Pass Needs the emergency line Toothache, Saturday Pass Asks for another patient's records Red team Pass Talks the agent into a discount Red team Fail ### Testing Simulated callers run every draft before it ships, including the ones trying to break it. Checks · last 7 days 1,284 calls Reads the date and time back 98.4% Never quotes treatment prices 99.7% Offers the next open slot 93.1% Transfers emergencies within two turns 81.2% Saturday emergencies wait on hold 14 calls · front desk line has no Saturday hours ### Monitoring Every live call scored against checks you write in plain language. ### Messaging Agents and your team answer texts from one inbox. ### Voices Pick from the library or clone a voice from a few clips. ### Every language A greeting, prompt and voice per language, switching mid-call. ### Realtime Run GPT-Live or Grok Voice end to end instead of a cascade. ## Router A hosted, provider-neutral STT, LLM and TTS data plane at router.speko.dev, with typed contracts and managed routing. ### Benchmark coverage by language Model EN (English) ES (Spanish) DE (German) FR (French) AR (Arabic) FIL (Filipino) NB (Norwegian) HI (Hindi) TA (Tamil) TE (Telugu) KO (Korean) ZH (Chinese (Mandarin)) JA (Japanese) TH (Thai) LT (Lithuanian) ID (Indonesian) SW (Swahili) deepgram:nova-3 soniox:stt-rt-v5 OpenAI:GPT-4o Transcribe bench openai:gpt-4o-mini-transcribe cartesia:ink-whisper google:chirp_3 Alibaba:Qwen3-ASR bench openai:gpt-transcribe xai:grok-stt assemblyai:universal-3-5-pro ElevenLabs:Scribe v2 bench Gladia:Solaria-1 bench Smallest AI:Pulse Pro bench ElevenLabs:Scribe v1 bench Alibaba:Qwen-ASR bench Deepgram:Nova-2 bench modulate:velma-2-stt-streaming azure:MAI-Transcribe-2 gradium:default AssemblyAI:Universal-2 bench Azure:Fast Transcription bench fish:transcribe-1 fish:transcribe-1-pro gemini:gemini-3.5-transcribe OpenAI:Whisper-1 bench alibaba:qwen3-asr-flash-realtime elevenlabs:scribe_v2_realtime meta:muse-voice-transcribe-1.0 AssemblyAI:Universal (batch tier) bench cartesia:ink-2 Groq:Whisper Large v3 bench hamsa:s3 hopper:nemotron-asr modulate:velma-2-stt-streaming-english-v2 munsit:munsit munsit:munsit-en-ar nari:qwen3-asr nari:qwen3-asr-fast smallest:pulse Hover a cell for its number. worse better 12 of 55 are only measured in English Their rank in any other language is unknown — including the model that sits at the top of the English table. 4 different models win across 17 languages No single model is best everywhere, so the right pick changes with the language your users speak. ### Score against cost, per stage Accuracy (WER · conversational) 25ms 1.75s 12.0% 24.2% Qwen3-ASR Fast 12.0% · 33ms Finalize (ms, p50) Model Qwen3-ASR Fast nari:qwen3-asr-fast 12.0% 33ms Qwen3-ASR nari:qwen3-asr 12.0% 25ms Qwen3-ASR Realtime alibaba:qwen3-asr-flash-realtime 12.2% 889ms Scribe v2 Realtime elevenlabs:scribe_v2_realtime 12.5% 1.75s Gemini 3.5 Transcribe Live gemini:gemini-3.5-transcribe-live 13.1% 1.28s GPT Live Transcribe openai:gpt-live-transcribe 13.5% 620ms Grok Voice Transcribe 1.0 xai:grok-stt 13.7% 1.13s Universal-3.6 Pro assemblyai:universal-3-6-pro 13.7% 505ms Muse Voice Transcribe meta:muse-voice-transcribe-1.0 14.2% — Velma 2 modulate:velma-2-stt-streaming-english-v2 14.4% 1.21s Pulse smallest:pulse 14.5% 839ms Universal-3.5 Pro assemblyai:universal-3-5-pro 14.5% 408ms stt-rt-v5 soniox:stt-rt-v5 16.6% 340ms Ink-2 cartesia:ink-2 16.7% 587ms GPT Realtime Whisper openai:gpt-realtime-whisper 16.8% 1.14s GPT Transcribe openai:gpt-transcribe 16.8% 817ms Nova-3 deepgram:nova-3 17.0% 189ms Chirp 3 google:chirp_3 18.4% — Flux deepgram:flux-general-en 19.7% 712ms GPT-4o-mini Transcribe openai:gpt-4o-mini-transcribe 20.6% 822ms flux-general-multi deepgram:flux-general-multi 21.4% 854ms nemotron-asr hopper:nemotron-asr 23.1% 656ms Gradium ASR gradium:default 24.2% 1.13s transcribe-1-pro fish:transcribe-1-pro — — transcribe-1 fish:transcribe-1 — — MAI-Transcribe-2 azure:MAI-Transcribe-2 — — MAI-Transcribe-2-Streaming azure:MAI-Transcribe-2-Streaming — — ink-whisper cartesia:ink-whisper — — gemini-3.5-transcribe gemini:gemini-3.5-transcribe — — s3 hamsa:s3 — — velma-2-stt-streaming modulate:velma-2-stt-streaming — — munsit munsit:munsit — — munsit-en-ar munsit:munsit-en-ar — — default palabra:default — — ## Gateway The open customer-side runtime for LiveKit and Pipecat: provider-direct streaming, local BYOK credentials and optional Speko-managed routes. ### Run and observe your voice workers Use the native Gateway integrations in your framework, or call the hosted Router through its public OpenAPI and AsyncAPI contracts. ```python from livekit.agents import AgentSession from speko_gateway.livekit import LLM, STT, TTS session = AgentSession( stt=STT(credential_source="auto"), llm=LLM(model="auto", objective="balanced"), tts=TTS(credential_source="auto"), ) ``` ## MCP Run Speko from Claude, ChatGPT, Grok, Cursor and the rest: the same agents, numbers and calls, driven from a chat or an editor. ### Your assistant can make the call Connect Speko to the AI tool you already work in. Build and deploy agents from a coding agent, then place calls and read transcripts from a chat. - [ClaudeAdd Speko from Claude's connector directory and place AI phone calls, read call transcripts and check your voice agents from a Claude chat. Set up Claude ›](https://speko.ai/integrations/claude) - [ChatGPTAdd the Speko app in ChatGPT to test voice agents, generate speech and place AI phone calls without leaving the chat. Set up ChatGPT ›](https://speko.ai/integrations/chatgpt) - [GrokInstall the Speko plugin in Grok Bot to place calls from your own number, in the US or abroad, and answer your business line with an AI receptionist. Set up Grok ›](https://speko.ai/integrations/grok) - [Claude CodeGive Claude Code the full Speko MCP surface: create and deploy voice agents, buy phone numbers, run evals and read call transcripts from your terminal. Set up Claude Code ›](https://speko.ai/integrations/claude-code) - [CursorAdd Speko's MCP server to Cursor and create, deploy and test voice agents next to the code that calls them. Set up Cursor ›](https://speko.ai/integrations/cursor) - [CodexConnect OpenAI Codex to Speko's MCP server to create, deploy and test voice agents and phone numbers from the CLI. Set up Codex ›](https://speko.ai/integrations/codex) - [ReplitConnect Replit Agent to Speko and ship an app with a voice agent and a phone number, built in the same session. Set up Replit ›](https://speko.ai/integrations/replit) Any other MCP client connects to mcp.speko.ai/mcp with OAuth or an API key. [All integrations](https://speko.ai/integrations) ## Point your agent at Speko ```bash $ claude mcp add --transport http speko https://mcp.speko.ai/mcp ``` [Read the docs](https://docs.speko.ai) [Get API key](https://platform.speko.ai/sign-in)