README.md

---
title: Qwen2.5-7B-Instruct Borg Merge v1
emoji: 🤖
colorFrom: indigo
colorTo: purple
sdk: gradio
sdk_version: 5.12.0
python_version: "3.12"
app_file: app.py
hardware: zero-a10g
pinned: false
license: apache-2.0
models:
  - Optitransfer/Qwen2.5-7B-Instruct-borg-merge-v1
short_description: "Chat with a 9-model cross-family merged checkpoint"
---

# Qwen2.5-7B-Instruct -- Borg Merge v1

Chat with a model created by merging **9 models from 4 architecture families** (Llama, Phi, NeoX, OPT) into a single checkpoint using training-free cross-family weight merging.

**Headline results**: +3.3 pp GSM8K, +3.2 pp ARC-Challenge, +2.6 pp IFEval over the unmerged anchor -- no fine-tuning and no distillation.

[Model card](https://huggingface.co/Optitransfer/Qwen2.5-7B-Instruct-borg-merge-v1) | [Paper (SSRN)](https://ssrn.com/abstract=6545518) | [crdt-merge](https://github.com/mgillr/crdt-merge) | [Deep-dive article](https://medium.com/@rgillespie83/we-merged-9-models-from-4-architecture-families-into-one-and-it-beats-the-anchor-on-real-e6537dfa9252)