whatmodel.sh / stt / Ultravox v0.6 (Llama 3.1 8B)

Ultravox v0.6 (Llama 3.1 8B)

+ contribute
Update model informationAdd a place to run itAdd a test resultPropose a benchmarkCorrect a benchmark scoreEdit data on GitHub

Ultravox v0.6 (Llama 3.1 8B) is an open-weights speech-to-text model from Fixie.ai (Ultravox), released June 2025. Licence: mit.

Catalogue identity notes

Exact checkpoint fixie-ai/ultravox-v0_6-llama-3_1-8b (CSV label 'ultravox/fixie-ai-ultravox-v0_6-llama-3_1-8b'). The repo holds 0.69B params (Whisper-large-v3-turbo encoder + projector); the Llama-3.1-8B-Instruct backbone (Llama 3.1 licence, gated) is loaded separately, so no total params_b is given. No local variant: the card offers only a chat-style trust_remote_code pipeline, not a verified transcription command.

48.5±5.1comparison only: no current run option verifiedfrom 1 benchmark result across 1 source* thinly tested, treat the rank as provisional

Score within the Open ASR English short-form WER (March 2026) comparison set only.

$ how it compares

The speech-to-text models ranked either side of Ultravox v0.6 (Llama 3.1 8B), out of 176 with a score.

#modelscorewhereRAMreleased
118OWSM-CTC v3.2 ft 1B ESPnet (CMU WAVLab)49±4local~5.4GB2024-09
119Nemotron Speech Streaming EN 0.6B NVIDIA49±4local~0.7GB2026-03
120Ultravox v0.6 (Llama 3.1 8B) Fixie.ai (Ultravox)49±5comparison only2025-06
121Whisper medium.en OpenAI48±3local~2.4GB2022-09
122Distil-Whisper distil-large-v2 Hugging Face (Distil-Whisper)48±2local~2.3GB2023-10

see all speech-to-text models · find the ones that fit your machine

$ what it's good at

other tests1 not in a task · 90% of its score
testresultvs the others in its rundifficultycountswhy
Open ASR English short-form WER (March 2026)8.5% ✎▁▁▁▃··▁▁▂██▆ beats 27/86avg 8.2%▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 90%blunt

$ where to run it

No verified API or download recorded yet · add one

Official weights and documentation: huggingface.co/fixie-ai ↗

$ how the score was worked out · Open ASR Leaderboard 90%

Each result implies a score on its own, after adjusting for how hard its test is and who ran it. The score is their weighted average, with a small pull toward the middle for thinly tested models. Each test's weight is its counts column above; here it's added up by who ran the tests. How it works

sourceresultsshare of the scorelicence
Open ASR Leaderboard1▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 90%Apache-2.0
prior · pull toward the middle▇▇▇ 10%
weighted average= 48.5

$ improve this page

Update model informationAdd a place to run itAdd a test resultPropose a benchmarkCorrect a benchmark scoreEdit data on GitHub

Use a short GitHub form or edit the data through a pull request. Every number needs a source; here's which ones we can use.

data updated · sources and licences · json