Athene-V2 is a 72B parameter model which excels at code completion, mathematics, and log extraction tasks.

tools 72b

Updated 1 year ago

1 year ago

f74199e5b2ed · 38GB ·

model

archqwen2

parameters72.7B

quantizationQ3_K_M

38GB

template

{{- if .Messages }} {{- if or .System .Tools }}<|im_start|>system {{- if .System }} {{ .System }} {{

1.5kB

license

Nexusflow.ai License Terms for Personal Use Release Date: 08/19/2024 "Agreement" means these terms a

6.8kB

system

You are Qwen, created by Alibaba Cloud. You are a helpful assistant.

68B

Readme

Athene-V2

Nexusflow’s Athene-V2 chat model, built on Qwen 2.5’s 72B foundation, achieves GPT-4o-level performance across key benchmarks while demonstrating how targeted optimization can enhance specific capabilities beyond traditional scaling approaches.

Model Features

72B parameters fine-tuned from Qwen 2.5
State-of-the-art chat performance matching or exceeding GPT-4o
Superior code completion (ranking #2 on bigcode-bench-hard)
Enhanced mathematics capabilities (MATH benchmark)
Precise long-form log extraction
Advanced post-training pipeline pushing the Pareto frontier

References

Blog post

HuggingFace