← sources

balbal-v2: Turkish speech recognition that runs on CPU, 2nd on Mu-Bench TR after Google Chirp 3

source: reddit (reddit:LocalLLaMA)published 2026-11-03 13:57ingested 2026-11-04 02:41

retained text

I've been working on Turkish ASR for a while and put out v2 recently. Two models, both CTC + KenLM, both meant for CPU (both trained for under $250):

\- balbal-v2-manas: 313M, Turkish + English, \~123x real time on 8 threads

\- balbal-v2: 45M, Turkish only, \~236x real time. It even runs faster than real time on one core of an old Celeron POS box running Windows 7.

Mu-Bench is a Turkish benchmark of 846 real customer service calls. Numbers from the official harness:

|Model|WER|sigWER|
|:-|:-|:-|
|Google Chirp 3|8.03|14.30|
|balbal-v2-manas|8.20|14.89|
|Gemini 3.5 Transcribe|8.68|18.20|
|ElevenLabs Scribe v2|9.43|19.15|
|balbal-v2|9.46|18.44|
|Deepgram Nova-3|13.37|24.35|
|Azure Speech|14.02|30.73|
|Qwen3-ASR 1.7B|14.76|30.61|
|whisper-large-v3-turbo|24.65|47.04|

On FLEURS Turkish manas is basically tied with whisper-large-v3-turbo (4.06 vs 4.07) but about 75x faster on the same CPU. English is ok too, 8.67 average WER on the Open ASR leaderboard sets.

The model uses a unique architecture that has a product-key-memory inspired by engrams and a refiner that loops 3 times before giving the output.

Also, turns out a team (LocalDoc) already trained an Azerbaijani model on the manas architecture, 10.98 WER on FLEURS Azerbaijani. (https://huggingface.co/LocalDoc/speakwrite-azerbaijani) I was happy that they took my model decisions which made me happy :D

\`pip install balbal\`, then \`balbal run file.wav\`. There's also a live mic mode and an OpenAI-compatible server.

Code is Apache-2.0. Weights are free for research, non-commercial use and small businesses, but you can't resell it as an ASR product or API. Full terms are on the model page.

Weights: [https://huggingface.co/erayyapagci/balbal-v2-manas](https://huggingface.co/erayyapagci/balbal-v2-manas) and [https://huggingface.co/erayyapagci/balbal-v2](https://huggingface.co/erayyapagci/balbal-v2)

Code: [https://github.com/erayyap/balbal](https://github.com/erayyap/balbal)

Demo + offline Android app: [https://erayy.com/balbal](https://erayy.com/balbal)

document fetched 2026-11-04 00:41 · canonical https://reddit.com/r/LocalLLaMA/comments/1xmtlm7/balbalv2_turkish_speech_recognition_that_runs_on