Cartesia 推出 Ink-2 語音轉文字模型,在 Artificial Analysis 串流排行榜以最低詞錯誤率排名第一
Introducing Ink-2: The #1-ranked STT built for voice agents
Cartesia 推出面向即時語音智能體的語音轉文字模型 Ink-2;截至 2026 年 6 月,Ink-2 在 Artificial Analysis 串流排行榜以最低詞錯誤率排名第一。
We’re excited to release Ink-2: a speech-to-text model built for real-time voice agents.
It’s ranked #1 on Artificial Analysis’s streaming leaderboard for lowest word error rate as of June 2026, with the most accurate built-in turn detection of any provider, so the model knows precisely when to listen and when to respond.
% of words transcribed incorrectly at Final Transcription after detected End of Speech. Lower is better.
June 2026
AA-AgentTalkVoxPopuliEarnings22
3.4%
2.4%
5.3%
Ink-2 (semantic endpoints)
Cartesia
2.8%
2.3%
6.4%
Scribe v2 Realtime
ElevenLabs
3.6%
2.1%
7.2%
U3.5 Pro Realtime
AssemblyAI
6.7%
4.9%
11.4%
Flux
Deepgram
Ink-2 (semantic endpoints)
Cartesia
3.4%
2.4%
5.3%
Scribe v2 Realtime
ElevenLabs
2.8%
2.3%
6.4%
U3.5 Pro Realtime
AssemblyAI
3.6%
2.1%
7.2%
Flux
Deepgram
6.7%
4.9%
11.4%
For voice agents, speech-to-text has to get three things right: accuracy, turn detection, and latency. If any one of these falls short, the experience breaks down. The agent may misunderstand the user, interrupt at the wrong time, respond too slowly, or make the conversation feel unnatural. Ink-2 was built to lead on all three.
Accuracy: Getting every word right
We’ve done extensive work on structured entity recognition like phone numbers, email addresses, alphanumerics, and dates. Ink-2 understands when it’s mid-entity and waits for the full sequence before committing, with no special prompting needed.
We built Ink-2 to be robust across a range of accents, which reflects real voice agent calls. On AppTek, a multi-accent benchmark spanning 14 English accents on real call-center dialogue, Ink-2 is the strongest streaming STT provider at 8% WER, vs. 10% for Deepgram Flux and 12% for ElevenLabs Scribe v2.
Multi-accent benchmark that measures WER across 14 English accents in real call-center dialogues. Lower is better.
8%
Ink-2
Cartesia
10%
Flux
Deepgram
12%
Scribe v2 Realtime
ElevenLabs
13%
U3.5 Pro Realtime
AssemblyAI
Accuracy also holds under production conditions, not just clean reference audio. Our internal benchmark samples audio directly from live voice agent calls covering non-native English speakers, background noise, and degraded audio from poor network conditions. Ink-2 achieves 6.5% WER, compared to 9.2% for ElevenLabs Scribe v2 and 9.4% for Deepgram Flux.
Benchmark built from internal dataset of 36 real calls with our own voice agents, captured in production conditions. Lower is better.
Cartesia evaluationJuly 2026
6.5%
Ink-2
Cartesia
9.2%
Scribe v2 Realtime
ElevenLabs
9.4%
Flux
Deepgram
10.7%
U3.5 Pro Realtime
AssemblyAI
Both structured entities and real-world production audio are where accuracy actually gets tested in a live voice agent, not just in a clean benchmark.
Turn detection: Knowing when to listen and when to respond
Accuracy shows up clearly in benchmarks, but turn detection is where most voice agents fall apart in production.
Most voice agents decide a turn is over based on silence. If someone pauses long enough, the turn abruptly ends. This works in a test environment, but on a real call it means cutting a customer off mid-address, or jumping in right after “and my email is…”
We created Ink-2 with built-in semantic endpointing: the model reads meaning, not silence, to decide when a turn is over. It knows when an incomplete address is still being given or when a trailing thought isn’t a stopping point. The turn stays open until the model is confident the speaker is done.
Ink-2 emits three events natively, with no external VAD needed:
turn.start— the user has begun speakingturn.eager_end— the model predicts the turn is wrapping up; your LLM can start generating earlyturn.end— turn confirmed complete
We measure turn detection against a human-labeled reference. Precision is how often the model is right when it calls a turn over (low precision means cutting people off), and recall is how often it catches every real end-of-turn (low recall means awkward dead air while the model keeps listening). F1 balances the two into a single score.
Ink-2 holds both precision and recall high at once, while the alternatives each trade one off for the other.
Internal benchmark that measures endpointing performance with live voice agent calls, captured in production conditions. Higher is better.
Cartesia evaluationJuly 2026
PrecisionRecallF1
89%
97%
93%
Ink-2
Cartesia
80%
97%
88%
Flux
Deepgram
74%
97%
84%
U3.5 Pro Realtime
AssemblyAI
89%
87%
88%
Scribe v2 Realtime
ElevenLabs
Ink-2
Cartesia
89%
97%
93%
Flux
Deepgram
80%
97%
88%
U3.5 Pro Realtime
AssemblyAI
74%
97%
84%
Scribe v2 Realtime
ElevenLabs
89%
87%
88%
At micro1, we use Ink-2 to power Zara, our AI recruiting agent that conducts voice interviews with candidates at scale. The latency is almost instant, and the turn-taking handles real conversational patterns really well.
Latency: Keeping the conversation flowing
Similar to TTS latency, transcription latency has a direct impact on how natural a conversation feels. The metric we care about is Time-to-Final-Transcript (TTFT): how long it takes to get a final transcript from the moment the user finishes speaking. This is what determines whether your agent feels like it’s paying attention or like it’s buffering.
Ink-2’s latency is lightning fast, with a TTFT of 0.1s. And because turn.eager_end lets your LLM get a head start before the turn is fully confirmed, your agent responds fast enough to feel like it’s actually in the conversation.
Join the teams powered by Ink-2
Ink-2 is live at play.cartesia.ai, available directly via API and across the platforms voice teams build on, including LiveKit, Vapi, and Pipecat.
We’re excited to keep building alongside the best teams pushing voice AI forward and running Ink-2 in production. We’ve already pushed out a world class English model, and multilingual support is on the way, so Ink-2 fits the way your users speak, wherever they are.
Try Ink-2
來源:Cartesia 產品與研究 · cartesia.ai