Qwen2.5-Coder (1.5B)

Training compute

5.1×10²² FLOP

Parameters

1.5B

Published

Sep 18, 2024

Qwen2.5-Coder (1.5B) is an AI model developed by Alibaba (China), first published in September 2024. It works in the language domain, on tasks such as code generation, code autocompletion, quantitative reasoning and 2 more.

Training it took an estimated 5.1×10²² FLOP of compute (estimation method: operation counting). The model has 1,540,000,000 parameters.

Access: Open weights (unrestricted). Its weights are openly available. Epoch AI rates the confidence of this record as confident.

Full record

Organization: Alibaba
Country of organization: China
Domain: Language
Task: Code generation, Code autocompletion, Quantitative reasoning, Question answering, Language modeling/generation
Training compute: 5.1×10²² FLOP
Compute estimation method: Operation counting
Parameters: 1,540,000,000
Model accessibility: Open weights (unrestricted)
Open weights: Yes
Epoch confidence: Confident

Benchmark results

01	Epoch Capabilities Index	ECI Score	102.7
02	GSM8K	EM	0.66
03	HellaSwag	Overall accuracy	0.62
04	WinoGrande	Accuracy	0.61
05	MMLU	EM	0.54
06	ARC AI2	Challenge score	0.45