durganani60/qwen2.5-0.5b-financial-adapted

TEXT GENERATIONPricing:Input $0.04 / Cached $0.008 / Output $0.08Concurrent Unit Cost:1Model Size:0.5BQuant:BF16Context Size:32kTool Calling:SupportedPublished:Sep 5, 2026License:apache-2.0Architecture:Transformer Open Weights Featherless Exclusive Cold

The durganani60/qwen2.5-0.5b-financial-adapted model is a 0.5 billion parameter causal language model, based on the Qwen2.5 architecture, developed by Durga. It is specifically adapted for financial NLP tasks, featuring a custom Byte-Level BPE Tokenizer to reduce fragmentation of financial terminology. This model excels at processing financial documents and generating domain-specific text with reduced prompt inflation, making it ideal for financial text generation and analysis.

Loading preview...

Model Overview

The durganani60/qwen2.5-0.5b-financial-adapted is a 0.5 billion parameter Causal Language Model, developed by Durga, and fine-tuned from Qwen/Qwen2.5-0.5B. Its core innovation lies in its domain adaptation for financial NLP tasks through a custom Byte-Level BPE Tokenizer. This tokenizer is integrated via Tokenizer Surgery and aligned using TRL SFTTrainer with PEFT/LoRA.

Key Capabilities

  • Specialized Tokenization: Addresses common issues in standard LLMs where financial terms (e.g., $NVDA, EBITDA, EUR/USD) are fragmented into excessive subwords. This reduces prompt inflation and [UNK] errors.
  • Vocabulary Extension: Extends the base vocabulary with specific financial domain tokens, resizing the embedding matrix and language model head accordingly.
  • Continued Pre-training: Utilizes LoRA to train new token embeddings while preserving the base model's general language capabilities.

Good For

  • Financial Text Generation: Creating and completing text within the financial domain.
  • Document Processing: Analyzing financial filings such as SEC 10-K, 10-Q, earnings transcripts, and analyst reports.
  • Domain-Specific Prompt Engineering: Achieving more efficient and accurate context handling for financial queries due to reduced token fragmentation.

Limitations

  • Domain Specificity: Primarily designed for financial text compression and processing, not general open-domain chat.
  • Accuracy Verification: Outputs for quantitative or regulatory compliance tasks require independent human verification.

Finetunes
1 models