Fluid is a speech-to-speech model built to solve problems. On τ²-bench it resolves 84% of support tasks. It speaks 24+ languages and 52 dialects. It talks like a person, delegates to the frontier model of your choice and works with OpenAI's Live Protocol.
Hi PH, I'm Kajo from KugelAudio. Alex built the model, here's a quick run down.
It's a speech-to-speech model built to solve problems. On τ²-bench, the customer-service benchmark, it resolves 84% of tasks, on par with other frontier models. It speaks 24+ languages and 52 dialects.
It also talks the way people talk. It backchannels while you speak (mhm, aha, etc.) and keeps you posted while it checks something. It's the most natural-sounding model we've built to date.
When the conversation needs a lookup, like checking an order, Fluid hands that step to the LLM you choose: ours, or the one you already use. It works with every voice we already support, and if you're on OpenAI's Live API, switching takes a new URL and key.
eu hosted speech to speech is a real wedge for regulated buyers 🇪🇺 latency under load ok?
The backchanneling detail is what caught me — most voice agents feel like they're waiting for their turn to talk, and the little "mhm, aha" cues are a big part of why real conversations don't feel like that. Curious how the handoff works in practice: when Fluid delegates a lookup to another model, does it stay in the loop and keep talking, or is there a noticeable pause while the other model takes over?
About Fluid on Product Hunt
“speech-to-speech built to solve problems”
Fluid launched on Product Hunt on November 16th, 2026 and earned 77 upvotes and 3 comments, placing #20 on the daily leaderboard. Fluid is a speech-to-speech model built to solve problems. On τ²-bench it resolves 84% of support tasks. It speaks 24+ languages and 52 dialects. It talks like a person, delegates to the frontier model of your choice and works with OpenAI's Live Protocol.
Fluid was featured in API (98.9k followers), Artificial Intelligence (484k followers) and Audio (2.2k followers) on Product Hunt. Together, these topics include over 157.4k products, making this a competitive space to launch in.
Who hunted Fluid?
Fluid was hunted by Kajo Kratzenstein. A “hunter” on Product Hunt is the community member who submits a product to the platform — uploading the images, the link, and tagging the makers behind it. Hunters typically write the first comment explaining why a product is worth attention, and their followers are notified the moment they post. Around 79% of featured launches on Product Hunt are self-hunted by their makers, but a well-known hunter still acts as a signal of quality to the rest of the community. See the full all-time top hunters leaderboard to discover who is shaping the Product Hunt ecosystem.
Want to see how Fluid stacked up against nearby launches in real time? Check out the live launch dashboard for upvote speed charts, proximity comparisons, and more analytics.
Hi PH, I'm Kajo from KugelAudio. Alex built the model, here's a quick run down.
It's a speech-to-speech model built to solve problems. On τ²-bench, the customer-service benchmark, it resolves 84% of tasks, on par with other frontier models. It speaks 24+ languages and 52 dialects.
It also talks the way people talk. It backchannels while you speak (mhm, aha, etc.) and keeps you posted while it checks something. It's the most natural-sounding model we've built to date.
When the conversation needs a lookup, like checking an order, Fluid hands that step to the LLM you choose: ours, or the one you already use. It works with every voice we already support, and if you're on OpenAI's Live API, switching takes a new URL and key.
Talk to it yourself and tell me where it breaks.