Backed by Y CombinatorThe Router for Voice AI
Every speech model, benchmarked language by language, wired into one API.
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
ES
DE
FR
AR
FIL
NB
HITATE
KO
ZH
JA
TH
LT
ID
SW
ENESDEFRARFILNBHITATEKOZHJATHLTIDSWdeepgram:nova-3
soniox:stt-rt-v5
OpenAI:GPT-4o Transcribebench
openai:gpt-4o-mini-transcribe
cartesia:ink-whisper
google:chirp_3
Alibaba:Qwen3-ASRbench
openai:gpt-transcribe
xai:grok-stt
assemblyai:universal-3-5-pro
ElevenLabs:Scribe v2bench
Gladia:Solaria-1bench
Smallest AI:Pulse Probench
ElevenLabs:Scribe v1bench
Alibaba:Qwen-ASRbench
Deepgram:Nova-2bench
modulate:velma-2-stt-streaming
azure:MAI-Transcribe-2
gradium:default
AssemblyAI:Universal-2bench
Azure:Fast Transcriptionbench
fish:transcribe-1
fish:transcribe-1-pro
gemini:gemini-3.5-transcribe
OpenAI:Whisper-1bench
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 v3bench
Hamsa:S3bench
modulate:velma-2-stt-streaming-english-v2
munsit:munsit
munsit:munsit-en-ar
nari:qwen3-asr
nari:qwen3-asr-fast
smallest:pulse
Score against cost, per stage
Accuracy (WER · conversational)
25ms1.75s12.0%24.2%
Qwen3-ASR Fast12.0% · 33ms
Finalize (ms, p50)
Model
Qwen3-ASR Fastnari:qwen3-asr-fast
12.0%33msQwen3-ASRnari:qwen3-asr
12.0%25msQwen3-ASR Realtimealibaba:qwen3-asr-flash-realtime
12.2%889msScribe v2 Realtimeelevenlabs:scribe_v2_realtime
12.5%1.75sGemini 3.5 Transcribe Livegemini:gemini-3.5-transcribe-live
13.1%1.28sGPT Live Transcribeopenai:gpt-live-transcribe
13.5%620msGrok Voice Transcribe 1.0xai:grok-stt
13.7%1.13sUniversal-3.6 Proassemblyai:universal-3-6-pro
13.7%505msMuse Voice Transcribemeta:muse-voice-transcribe-1.0
14.2%—Velma 2modulate:velma-2-stt-streaming-english-v2
14.4%1.21sPulsesmallest:pulse
14.5%839msUniversal-3.5 Proassemblyai:universal-3-5-pro
14.5%408msstt-rt-v5soniox:stt-rt-v5
16.6%340msInk-2cartesia:ink-2
16.7%587msGPT Realtime Whisperopenai:gpt-realtime-whisper
16.8%1.14sGPT Transcribeopenai:gpt-transcribe
16.8%817msNova-3deepgram:nova-3
17.0%189msChirp 3google:chirp_3
18.4%—Fluxdeepgram:flux-general-en
19.7%712msGPT-4o-mini Transcribeopenai:gpt-4o-mini-transcribe
20.6%822msflux-general-multideepgram:flux-general-multi
21.4%854msGradium ASRgradium:default
24.2%1.13stranscribe-1-profish:transcribe-1-pro
——transcribe-1fish:transcribe-1
——MAI-Transcribe-2azure:MAI-Transcribe-2
——MAI-Transcribe-2-Streamingazure:MAI-Transcribe-2-Streaming
——ink-whispercartesia:ink-whisper
——gemini-3.5-transcribegemini:gemini-3.5-transcribe
——velma-2-stt-streamingmodulate:velma-2-stt-streaming
——munsitmunsit:munsit
——munsit-en-armunsit:munsit-en-ar
——defaultpalabra: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.
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"),
)Point your agent at Speko
$claude mcp add --transport http speko https://mcp.speko.ai/mcp