惯性聚合 高效追踪和阅读你感兴趣的博客、新闻、科技资讯
阅读原文 在惯性聚合中打开

推荐订阅源

罗磊的独立博客
美团技术团队
Apple Machine Learning Research
Apple Machine Learning Research
Hugging Face - Blog
Hugging Face - Blog
让小产品的独立变现更简单 - ezindie.com
让小产品的独立变现更简单 - ezindie.com
月光博客
月光博客
WordPress大学
WordPress大学
The Cloudflare Blog
阮一峰的网络日志
阮一峰的网络日志
freeCodeCamp Programming Tutorials: Python, JavaScript, Git & More
博客园_首页
博客园 - Franky
博客园 - 司徒正美
酷 壳 – CoolShell
酷 壳 – CoolShell
爱范儿
爱范儿
Jina AI
Jina AI
Last Week in AI
Last Week in AI
雷峰网
雷峰网
IT之家
IT之家
OSCHINA 社区最新新闻
OSCHINA 社区最新新闻
博客园 - 聂微东
小众软件
小众软件
奇客Solidot–传递最新科技情报
奇客Solidot–传递最新科技情报
V
V2EX

OpenAI News

Using custom GPTs ChatGPT for customer success teams Applications of AI at OpenAI Research with ChatGPT Analyzing data with ChatGPT Financial services Responsible and safe use of AI Writing with ChatGPT ChatGPT for research Creating images with ChatGPT Personalizing ChatGPT ChatGPT for finance teams Getting started with ChatGPT Working with files in ChatGPT Learn ChatGPT workflows for sales teams Prompting fundamentals ChatGPT for managers Using projects in ChatGPT Learn ChatGPT workflows for marketing teams Brainstorming with ChatGPT AI fundamentals ChatGPT for operations teams Healthcare Our response to the Axios developer tool compromise Using skills OpenAI Full Fan Mode Contest: Terms & Conditions CyberAgent moves faster with ChatGPT Enterprise and Codex The next phase of enterprise AI 儿童安全蓝图正式发布 推出 OpenAI 安全研究员计划
OpenAI o1-mini
2024-09-12 · via OpenAI News
OpenAI

Advancing cost-efficient reasoning.

We're releasing OpenAI o1‑mini, a cost-efficient reasoning model. o1‑mini excels at STEM, especially math and coding—nearly matching the performance of OpenAI o1 on evaluation benchmarks such as AIME and Codeforces. We expect o1‑mini will be a faster, cost-effective model for applications that require reasoning without broad world knowledge.

Today, we are launching o1‑mini to tier 5 API users(opens in a new window) at a cost that is 80% cheaper than OpenAI o1‑preview. ChatGPT Plus, Team, Enterprise, and Edu users can use o1‑mini as an alternative to o1‑preview, with higher rate limits and lower latency (see Model Speed).

Large language models such as o1 are pre-trained on vast text datasets. While these high-capacity models have broad world knowledge, they can be expensive and slow for real-world applications. In contrast, o1‑mini is a smaller model optimized for STEM reasoning during pretraining. After training with the same high-compute reinforcement learning (RL) pipeline as o1, o1‑mini achieves comparable performance on many useful reasoning tasks, while being significantly more cost efficient.

When evaluated on benchmarks requiring intelligence and reasoning, o1‑mini performs well compared to o1‑preview and o1. However, o1‑mini performs worse on tasks requiring non-STEM factual knowledge (see Limitations).

Math Performance vs Inference Cost

AIMEInference Cost (%)

Mathematics: In the high school AIME math competition, o1‑mini (70.0%) is competitive with o1 (74.4%)–while being significantly cheaper–and outperforms o1‑preview (44.6%). o1‑mini’s score (about 11/15 questions) places it in approximately the top 500 US high-school students.

Coding: On the Codeforces competition website, o1‑mini achieves 1650 Elo, which is again competitive with o1 (1673) and higher than o1‑preview (1258). This Elo score puts the model at approximately the 86th percentile of programmers who compete on the Codeforces platform. o1‑mini also performs well on the HumanEval coding benchmark and high-school level cybersecurity capture the flag challenges (CTFs).

Codeforces

90012581650Elo

HumanEval

90.2%92.4%92.4%Accuracy

Cybersecurity CTFs

20.0%43.0%28.7%Accuracy (Pass@12)

STEM: On some academic benchmarks requiring reasoning, such as GPQA (science) and MATH-500, o1‑mini outperforms GPT‑4o. o1‑mini does not perform as well as GPT‑4o on tasks such as MMLU and lags behind o1‑preview on GPQA due to its lack of broad world knowledge.

MMLU
0-shot CoT

92.3%90.8%85.2%88.7%

GPQA
Diamond, 0-shot CoT

77.3%73.3%60.0%53.6%

MATH-500
0-shot CoT

94.8%85.5%90.0%60.3%

Human preference evaluation: We had human raters compare o1‑mini to GPT‑4o on challenging, open-ended prompts in various domains, using the same methodology as our o1‑preview vs GPT‑4o comparison. Similar to o1‑preview, o1‑mini is preferred to GPT‑4o in reasoning-heavy domains, but is not preferred to GPT‑4o in language-focused domains.

DomainWin Rate vs GPT-4o (%)

As a concrete example, we compared responses from GPT‑4o, o1‑mini, and o1‑preview on a word reasoning question. While GPT‑4o did not answer correctly, both o1‑mini and o1‑preview did, and o1‑mini reached the answer around 3-5x faster.

Chat speed comparison

o1‑mini is trained using the same alignment and safety techniques as o1‑preview. The model has 59% higher jailbreak robustness on an internal version of the StrongREJECT dataset compared to GPT‑4o. Before deployment, we carefully assessed the safety risks of o1‑mini using the same approach to preparedness, external red-teaming, and safety evaluations as o1‑preview. We are publishing the detailed results from these evaluations in the accompanying system card.

Metric

GPT-4o

o1-mini

% Safe completions refusal on harmful prompts (standard)

0.99

0.99

% Safe completions on harmful prompts (Challenging: jailbreaks & edge cases)

0.714

0.932

% Compliance on benign edge cases (“not over-refusal”)

0.91

0.923

Goodness@0.1 StrongREJECT jailbreak eval  (Souly et al. 2024(opens in a new window))

0.22

0.83

Human sourced jailbreak eval

0.77

0.95

Due to its specialization on STEM reasoning capabilities, o1‑mini’s factual knowledge on non-STEM topics such as dates, biographies, and trivia is comparable to small LLMs such as GPT‑4o mini. We will improve these limitations in future versions, as well as experiment with extending the model to other modalities and specialities outside of STEM.