# Upstage — Upstage 13B LLM Beats MS ToRA and ChatGPT in Mathematical Reasoning

- Company: Upstage (upstage.ai)
- Announced: 2024-01-08
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://www.upstage.ai/news/upstages-13b-llm
- Record: https://forck.live/items/18449-upstage-13b-llm-beats-ms-tora-and-chatgpt-in-mathematical-reasoning
- Subject: Solar Pro / Solar Mini
- Models affected: Upstage-Qanda 13B

Upstage announced that its 13B-parameter model, built with Qanda and KT, achieved state-of-the-art results on GSM8K and MATH benchmarks, surpassing MS ToRA 13B and ChatGPT, and outperforming GPT-4 on MATH with 48.8% accuracy. The model integrates natural language reasoning with code-based calculation. Further details are to be published in a research paper.

## Evidence

Verbatim from https://www.upstage.ai/news/upstages-13b-llm:

> The result: Upstage-Qanda 13B model has achieved SoTA performance over MS ToRA 13B on both GSM8K and MATH benchmark datasets. Notably, the model has surpassed ChatGPT's average performance across various benchmark tests and even outperformed GPT-4 on MATH with an impressive accuracy of 48.8 percent, showcasing its competitiveness with industry-leading models.

---

Record: https://forck.live/items/18449-upstage-13b-llm-beats-ms-tora-and-chatgpt-in-mathematical-reasoning
Catalogue: https://forck.live/llms.txt
Current issue: https://forck.live/feed.md