nassimjp/LFM2.5-1.2B-Instruct-Pashto

TEXT GENERATIONPricing:Input $0.04 / Cached $0.008 / Output $0.08Concurrent Unit Cost:1Model Size:1.2BQuant:BF16Context Size:32kPublished:Jul 27, 2026License:otherArchitecture:Transformer Featherless Exclusive Cold

The nassimjp/LFM2.5-1.2B-Instruct-Pashto is a 1.2 billion parameter instruction-tuned language model, building upon the LiquidAI/LFM2.5-1.2B-Base architecture. This model is specifically designed and optimized for the Pashto language, incorporating a specialized tokenizer for Pashto atomic units. It is intended for applications requiring high-quality Pashto language understanding and generation, leveraging supervised fine-tuning on Pashto instruction datasets.

Loading preview...

Model Overview

The nassimjp/LFM2.5-1.2B-Instruct-Pashto is a 1.2 billion parameter instruction-tuned language model, derived from the LiquidAI/LFM2.5-1.2B-Base architecture. Its primary distinction lies in its specialized focus on the Pashto language, incorporating a custom tokenizer designed to handle unique Pashto atomic units such as "ښ", "څ", "ځ", and others.

Key Capabilities

  • Pashto Language Processing: Optimized for understanding and generating text in Pashto.
  • Instruction Following: Fine-tuned with high-quality Pashto instruction datasets to respond to prompts effectively.
  • Specialized Tokenization: Utilizes a tokenizer specifically adapted for the nuances of the Pashto script, enhancing linguistic accuracy.

Training Methodology

The model's development involved:

  • Continued Pretraining (CPT): Requires extensive, deduped, and domain-balanced Pashto corpora.
  • Supervised Fine-Tuning (SFT): Leverages high-quality Pashto instruction datasets formatted for chat applications.

Use Cases

This model is particularly well-suited for applications requiring robust Pashto language capabilities, including:

  • Pashto-specific chatbots and conversational AI.
  • Text generation and summarization in Pashto.
  • Language understanding tasks for Pashto content.

Licensing

The model operates under the same license terms as its base model, LiquidAI/LFM2.5-1.2B-Base. Users should consult the original model card for detailed licensing information.