Atlas

Models

← All models

Qwen2.5 0.5B Instruct

Alibaba · 2024-09-19 · 494M parameters

Qwen2.5-0.5B-Instruct is the instruction-tuned version of Alibaba's Qwen2.5-0.5B, released on September 19, 2024. It is a dense decoder-only Transformer with 494,032,768 parameters, grouped-query attention and tied embeddings, with a 32,768-token context window and up to 8,192 generated tokens.

Benchmark scores

BenchmarkScore
Capability77.6
GPQA - release-reported, subset unspecified29.8
GSM8K49.6
HumanEval35.4
IFEval (prompt-level strict accuracy)27.9
LiveBench 2024-08-3112.6
LiveCodeBench - May 2023 to September 20245.1
MATH34.4
MBPP49.6
MMLU Pro15.0
MMLU-Redux24.1
MultiPL-E (HumanEval, 8-language average)28.5
Loading Atlas data…