newmindai/Llama-3.1-8B-Instruct-w16a8-1node-bs8

TEXT GENERATIONPricing:Input $0.2 / Cached $0.028 / Output $0.32Concurrent Unit Cost:1Model Size:8BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Oct 30, 2025License:llama3.1Architecture:Transformer Featherless Exclusive Cold

The newmindai/Llama-3.1-8B-Instruct-w16a8-1node-bs8 is an 8 billion parameter Llama 3.1-based instruction-tuned model, specifically optimized for Turkish legal domain tasks. It was trained by newmindai using Float8 mixed precision (FP8 activations, BF16 weights) within an FSDP2 distributed training framework to evaluate training speed and memory efficiency. This model is a variant of Llama-3.1-8B-Instruct, fine-tuned for legal-domain performance with a 32768 token context length.

Loading preview...

Overview

This model, newmindai/Llama-3.1-8B-Instruct-w16a8-1node-bs8, is an 8 billion parameter instruction-tuned variant of Meta's Llama-3.1-8B-Instruct, specifically adapted for the Turkish legal domain. It was developed by newmindai as part of an experiment comparing FSDP2 with bfloat16 precision against FSDP2 with FP8 mixed precision (bfp16-fp8) to assess the impact on training speed, memory efficiency, and legal-domain performance.

Key Characteristics

  • Base Architecture: Llama 3.1 with 8 billion parameters and a 32,768 token context length.
  • Precision: Trained using Float8 mixed precision (FP8 for activations, BF16 for weights) with Tensorwise quantization scaling.
  • Training Framework: Utilizes FSDP2 (Fully Sharded Data Parallel) for distributed training on a single node with 4 NVIDIA H100 GPUs.
  • Domain Specificity: Instruction-tuned for Turkish legal tasks, leveraging the EuroHPC-Legal dataset.

Use Cases

This model is primarily intended for:

  • Turkish Legal Text Processing: Answering questions, summarizing, or generating text related to Turkish legal documents.
  • Research in Mixed Precision Training: Evaluating the practical application and performance of FP8 mixed precision in large language model training.

Ethical Considerations

It is crucial to note that this model is for research and development purposes only and should not be used as a substitute for professional legal counsel. Users must ensure compliance with data protection and sector-specific regulations, and be aware of potential biases in the training data and model outputs.

Finetunes
1 models