Atlas

Models

← All models

Qwen2.5 1.5B Instruct

Alibaba · 2024-09-19 · 1.5B parameters

Qwen2.5-1.5B-Instruct is the instruction-tuned version of Alibaba's Qwen2.5-1.5B, released on September 19, 2024. It is a dense decoder-only Transformer with 1,543,714,304 parameters, grouped-query attention and tied embeddings, with a 32,768-token context window and up to 8,192 generated tokens.

Benchmark scores

BenchmarkScore
Capability99.2
GPQA - release-reported, subset unspecified29.8
GSM8K73.2
HumanEval61.6
IFEval (prompt-level strict accuracy)42.5
LiveBench 2024-08-3118.8
LiveCodeBench - May 2023 to September 202414.8
MATH55.2
MBPP63.2
MMLU Pro32.4
MMLU-Redux50.7
MultiPL-E (HumanEval, 8-language average)50.4
Loading Atlas data…