Open models behind BültenHub (bultenhub.com) and the BHAI demo.
AI & ML interests
Turkish news NLP: news writing, text classification, event clustering
Recent Activity
Organization Card
BültenHub
Newsroom automation for Turkish local publishers · bultenhub.com
BültenHub monitors Turkish news sources around the clock, groups stories by event and writes original, source-linked news with its own models. It then turns them into bulletins, voice-overs and news videos.
Open models (CC BY-NC 4.0)
| Model | What it does |
|---|---|
| BHAI | Qwen3.5-4B LoRA that writes a headline and a news text from a Turkish source. Runs in production on llama.cpp. |
| BHKAI | Fast Turkish news category classifier (TF-IDF + SGD). Retrained nightly, promoted only if it passes its exam. |
| BHDAI | Same-event matcher that groups stories from different sources. Retrained nightly, promoted only if it passes its exam. |
Latest versions
Models are retrained at night and published only after passing their exam. Updated 2026-11-18.
| Model | Version | Date | Exam |
|---|---|---|---|
| BHAI | qwen35-4b-hoca-mini2-20261031 | 2026-10-31 | exam A score 0.852, grounding 0.986, coverage 0.743 |
| BHKAI | bhkai-20261116-alpha1e-6 | 2026-11-16 | at confidence ≥ 0.8: coverage 57.6% (was 52.9%), agreement with teacher 98.1% (was 97.9%) |
| BHDAI | bhdai-20261118 | 2026-11-18 | at threshold 0.1: same-event pairs passed without a call 68.0% (was 68.4%), hard negatives wrongly passed 2.73% (was 2.73%) |
Changelog
- 2026-11-18 BHDAI
bhdai-20261118: at threshold 0.1: same-event pairs passed without a call 68.0% (was 68.4%), hard negatives wrongly passed 2.73% (was 2.73%) - 2026-11-17 BHDAI
bhdai-20261117: at threshold 0.1: same-event pairs passed without a call 67.8% (was 68.0%), hard negatives wrongly passed 2.75% (was 2.74%) - 2026-11-16 BHKAI
bhkai-20261116-alpha1e-6: at confidence ≥ 0.8: coverage 57.6% (was 52.9%), agreement with teacher 98.1% (was 97.9%) - 2026-11-16 BHDAI
bhdai-20261116: at threshold 0.1: same-event pairs passed without a call 68.4% (was 68.5%), hard negatives wrongly passed 3.61% (was 3.65%) - 2026-11-15 BHKAI
bhkai-20261115: at confidence ≥ 0.8: coverage 51.4% (was 51.4%), agreement with teacher 95.9% (was 95.6%) - 2026-11-15 BHDAI
bhdai-20261115: at threshold 0.1: same-event pairs passed without a call 69.7% (was 69.4%), hard negatives wrongly passed 3.45% (was 3.37%) - 2026-11-14 BHDAI
bhdai-20261114: at threshold 0.1: same-event pairs passed without a call 69.3% (was 69.3%), hard negatives wrongly passed 2.64% (was 2.64%) - 2026-11-13 BHKAI
bhkai-20261113: at confidence ≥ 0.8: coverage 50.2% (was 49.7%), agreement with teacher 96.4% (was 95.3%) - 2026-11-13 BHDAI
bhdai-20261113: at threshold 0.1: same-event pairs passed without a call 69.3% (was 69.0%), hard negatives wrongly passed 2.64% (was 2.67%) - 2026-11-12 BHKAI
bhkai-20261112: at confidence ≥ 0.8: coverage 48.1% (was 47.8%), agreement with teacher 96.9% (was 96.5%) - 2026-11-12 BHDAI
bhdai-20261112-c03: at threshold 0.1: same-event pairs passed without a call 69.2% (was 69.7%), hard negatives wrongly passed 3.14% (was 3.22%) - 2026-11-11 BHDAI
bhdai-20261111: at threshold 0.1: same-event pairs passed without a call 69.1% (was 68.9%), hard negatives wrongly passed 2.61% (was 2.61%)
Try it
BHAI demo: paste your own notes and get a Turkish news draft (free ZeroGPU).
🇹🇷 BültenHub, Türkiye'deki yerel yayıncılar için haber otomasyonu platformu. Modellerimiz açık; geri bildirimlerinizi bekliyoruz.
datasets 0
None public yet