Zrald/zraldaiv1.1
Zrald/zraldaiv1.1 is an 8 billion parameter causal language model developed by Zrald, serving as a compressed variant of Qwen3-8B. This model offers a balance between performance and efficiency, achieving a Wikitext-2 perplexity of 10.40 and 100% on MMLU-mini. It is suitable for applications requiring a smaller footprint while maintaining strong language understanding capabilities.
Loading preview...
ZraldAI v1.1 Overview
Zrald/zraldaiv1.1 is an 8 billion parameter causal language model, developed by Zrald, designed as a compressed version of the Qwen3-8B architecture. This model aims to provide efficient performance for various natural language processing tasks.
Key Characteristics
- Architecture: A compressed variant of Qwen3-8B.
- Parameter Count: 8 billion parameters.
- Context Length: Supports a context length of 32,768 tokens.
Performance Metrics
While being a compressed model, Zrald/zraldaiv1.1 demonstrates competitive performance:
- Wikitext-2 Perplexity: Achieves a perplexity score of 10.40, compared to Qwen3-8B's 9.50.
- MMLU-mini: Scores 100% on the MMLU-mini benchmark, matching Qwen3-8B's performance.
When to Use This Model
This model is particularly well-suited for use cases where:
- Resource Efficiency is Key: Its compressed nature makes it a good choice for environments with limited computational resources.
- Strong Language Understanding is Required: Despite compression, it maintains high performance on benchmarks like MMLU-mini.
- Qwen3-8B Compatibility is Desired: As a variant of Qwen3-8B, it may offer similar characteristics in a more compact form.