P0x0/Astra-v1-12B

TEXT GENERATIONPricing:Input $1.2 / Output $4.8Concurrent Unit Cost:1Model Size:12.2BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Sep 21, 2024License:apache-2.0Architecture:Transformer0.0K Open Weights Featherless Exclusive Cold

Astra-v1-12B is a 12.2 billion parameter general-purpose transformer-based language model developed by P0x0, fine-tuned from Mistral-Nemo-Base-2407. It is specifically optimized to replicate the high-quality generation style of Claude 3's Sonnet and Opus models. This model excels at instruction-following tasks, making it suitable for text generation, summarization, and question answering.

Loading preview...

Overview

Astra-v1-12B is a 12.2 billion parameter language model developed by P0x0, fine-tuned from the Mistral-Nemo-Base-2407 architecture. Its primary differentiation lies in its fine-tuning objective: to emulate the generation quality and style of Claude 3's Sonnet and Opus models. This makes Astra-v1-12B particularly adept at instruction-following tasks across various natural language processing applications.

Key Capabilities

  • Instruction Following: Designed to respond to prompts with a quality and style similar to Claude 3 models.
  • General NLP Tasks: Proficient in text generation, summarization, and question answering.
  • Dialogue Systems: Can be integrated into conversational AI applications.

Performance Metrics

The model's performance is evaluated across several benchmarks, including:

  • IFEval: 28.06 (inst_level_strict_acc, prompt_level_strict_acc)
  • BBH: 31.81 (acc_norm)
  • MMLU-PRO: 27.34 (acc)

Good For

  • Developers seeking a model with a Claude 3-like generation style for general NLP tasks.
  • Applications requiring robust text generation, summarization, or question answering capabilities.
  • Experimentation with instruction-tuned models for diverse use cases.

Finetunes
3 models