FluidVoice’s new Fluid Speech Pro model tops the Open ASR benchmark
By Steven Van ·
FluidVoice says Fluid Speech Pro ranks first on the Open ASR benchmark. It is live in the FluidVoice beta, with open weights coming soon.
FluidVoice introduced Fluid Speech Pro on October 25. FluidVoice says the new speech recognition model is number one on the Open ASR benchmark across nine public sets.
FluidVoice’s chart gives Fluid Speech Pro an average word error rate of 4.88, where lower is better. That puts it just ahead of Bark 3.0 Pro (4.93) and Qwen3-ASR-1.7B (4.94), and well ahead of Whisper large-v3 (6.66). Fluid Speech Mini, released on October 20, scores 5.31.
FluidVoice says the model is still early. You can try it now in the FluidVoice beta, and open weights are coming soon. The beta also has Fluid 1.1 Splash, a multilingual research preview that FluidVoice calls its best cleanup model yet for formatting your dictation. FluidVoice hasn’t shared Speech Pro’s size or license.
Sources: FluidVoice’s announcement on X.