Qwen 2

Qwen2 is a series of state of the art, multi-lingual LLMs that range from 0.5 to 72 billion parameters. It excels in a variety of tasks. The only variant offered by Rubra at the moment is the 7B one.

note

Model	Params	Context Length	GQA	Token Count	Knowledge Cutoff
Qwen2 7B Instruct	7.6B	131,072	Yes	3T	2023

tip

The Rubra enhanced Qwen2-7b-Instruct model is capable of doing tool/function calling in Chinese!

We recommend this model for any task exceeding the context limit found in Llama-3 models.

Qwen2 7B Instruct

• rubra-ai/Qwen2-7B-Instruct
• rubra-ai/Qwen2-7B-Instruct-GGUF

MT-bench Pairwise Comparison:

Model	Function Calling	MMLU	GPQA	GSM-8K	MATH	MT-bench	MT-bench Pairwise Comparison
							Win	Loss	Tie	Win Rate	Loss Rate	Adjusted Win Rate
Qwen2-7B-Instruct	-	70.78	32.14	78.54	30.10	8.29	49	33	78	0.30625	0.20625	0.55
Rubra Enhanced Qwen2-7B-Instruct	85.71%	68.88	30.36	75.82	28.72	8.08	33	49	78	0.20625	0.30625	0.45