Mistral

Mistral is a series of open weight models that set the bar for efficiency, and are available for free under Apache 2.0 that allows use of the models anywhere without any restriction. The Mistral-7b-instruct-v0.3 model offers 32k context length, making it an ideal LLM for scenarios that require longer context.

note

Model	Params	Context Length	GQA	Token Count	Knowledge Cutoff
Mistral 7B Instruct v0.3	7B	32k	Yes	Undisclosed	~February 2023
Mistral 7B Instruct v0.2	7B	32k	Yes	Undisclosed	~February 2023

Mistral 7B Instruct v3

• rubra-ai/Mistral-7B-Instruct-v0.3
• rubra-ai/Mistral-7B-Instruct-v0.3-GGUF

Model	Function Calling	MMLU	GPQA	GSM-8K	MATH	MT-bench	MT-bench Pairwise Comparison
							Win	Loss	Tie	Win Rate	Loss Rate	Adjusted Win Rate
Mistral 7B Instruct v0.3	-	62.10	30.58	53.07	12.98	7.50	34	54	72	0.2125	0.3375	0.4375
Rubra Enhanced Mistral 7B Instruct v0.3	73.57%	59.12	29.91	43.29	11.14	7.69	54	34	72	0.3375	0.2125	0.5625

Mistral 7B Instruct v2

• rubra-ai/Mistral-7B-Instruct-v0.2
• rubra-ai/Mistral-7B-Instruct-v0.2-GGUF

Model	Function Calling	MMLU	GPQA	GSM-8K	MATH	MT-bench	MT-bench Pairwise Comparison
							Win	Loss	Tie	Win Rate	Loss Rate	Adjusted Win Rate
Mistral 7B Instruct v0.2	-	59.27	27.68	43.21	10.30	7.50	34	54	72	0.2125	0.3375	0.4375
Rubra Enhanced Mistral 7B Instruct v0.2	69.28%	58.90	29.91	34.12	8.36	7.36	59	43	58	0.36875	0.26875	0.55