uran1um1/tenorio-0.6b

Tenorio 0.6b: Spanish Language Optimized Model

Tenorio 0.6b is a compact, 0.8 billion parameter model developed by uran1um1, fine-tuned from Alibaba's Qwen3 0.6b. Its primary focus is to deliver superior performance in the modern Spanish language, making it suitable for applications where high-quality Spanish output is crucial.

Key Capabilities and Training:

• Enhanced Spanish Proficiency: The model has undergone continued pretraining on over 100 public domain Spanish books and instruction tuning with 7077 Spanish Q&A pairs.
• Improved Output Quality: It exhibits better context awareness, reduced hallucination, and more concise and natural answers when responding in Spanish.
• Reasoning Destruction: Deliberately designed to minimize reasoning abilities, which helps in reducing hallucinations and preventing infinite loops, leading to more direct and accurate Spanish responses.

Benchmarks:

While developed as a hobby project, custom benchmarks indicate its effectiveness in Spanish language quality. Using a GPT-OSS 20b model to judge grammar and language quality, Tenorio 0.6b showed improved performance over its base model in various categories:

• Creative writing: 40/80
• General STEM: 42/80
• Coding & CS: 45/80

These scores reflect its ability to produce grammatically sound and high-quality Spanish text, rather than content accuracy or reasoning.

Good For:

• Applications requiring precise and natural Spanish language generation.
• Use cases where complex reasoning is not needed, prioritizing clear and grammatically correct Spanish communication.
• Edge or tiny model deployments due to its compact size.