← Catalog
Ayb-0.5-0.8B
AvailableMetricOpen weights
Smallest model in Metric's Ayb 0.5 family of Armenian-language models, released Nov 20 2026. Built on Qwen3.5-0.8B-Base with an Armenian-extended tokenizer, continued pretraining on Armenian and English text, instruction tuning and initial post-training. Aimed at lightweight Armenian text processing and as a language component in speech, TTS and OCR pipelines; ranks 16th on ArmBench (0.5300, vendor-reported). Weights on Hugging Face under CC BY-NC 4.0 (non-commercial, gated with manual approval). Context length not stated.
Specifications
- License
- Open weights · CC-BY-NC-4.0
- Weights
- Downloadable
- Architecture
- Dense
- Parameters
- 0.8B
- Context window
- — tokens
- Max output
- —
- Knowledge cutoff
- —
- Price (in / out, $/M)
- —
- Modalities
- TextVision
Benchmarks
No benchmark scores recorded yet. Spotted some? Submit a correction.
Vendor-reported figures are claims until independently verified. See methodology.