Orca 2 7B

Orca 2 7B is a large language model developed by Microsoft, focusing on reasoning tasks and providing precise single-turn responses. It is a fine-tuned version of the LLaMA-2 architecture, trained on a synthetic dataset with enhanced reasoning capabilities, moderated by Microsoft Azure content filters. While adept at handling reasoning over user-provided data, reading comprehension, math problem-solving, and text summarization, it is not optimized for chat applications without further fine-tuning. Orca 2 shows strong performance in zero-shot settings but shares some LLMs' common limitations, including biases and the potential for generating misleading content. Designed primarily for research, its use in production requires careful assessment to mitigate potential harms or biases.

Orca 2 7B is an open-weight model in the Orca 2 family. The structured metadata tracks a 4k-token context window. This page tracks provider routes through Microsoft Foundry, with the cheapest tracked route listed at $0.52 input and $0.67 output per 1M tokens. Headline tracked benchmarks include HumanEval 28.4 and Massive Multitask Language Understanding 66.5.