01-ai/Yi-1.5-34B

Yi-1.5-34B Overview

Yi-1.5-34B is a 34 billion parameter model from 01-ai, representing an upgraded iteration of the original Yi series. It has undergone continuous pre-training with an additional 500 billion high-quality tokens and extensive fine-tuning on 3 million diverse samples.

Key Capabilities

• Enhanced Performance: Delivers stronger capabilities in coding, mathematical problem-solving, and general reasoning compared to its predecessor.
• Instruction Following: Shows improved ability to follow complex instructions accurately.
• Core Language Skills: Retains excellent performance in language understanding, commonsense reasoning, and reading comprehension.
• Context Length: Supports context lengths up to 32K tokens, allowing for processing longer inputs and generating more coherent extended outputs.

Benchmarks

• The Yi-1.5-34B base model demonstrates performance on par with or exceeding larger models in various benchmarks.
• The Yi-1.5-34B-Chat variant is noted for being competitive with or outperforming larger models across most benchmarks.

Good For

• Applications requiring robust coding and mathematical reasoning.
• Tasks demanding strong instruction-following and complex problem-solving.
• General language understanding and generation where high accuracy and coherence are critical.