公司指南

OpenAI o1 和 o3 推理模型解釋

OpenAI o1 和 o3 是推理模型,它們在回答之前使用額外的測試時計算來解決數學、科學和編碼中的多步驟問題。

聆聽此指南AI 語音敘述 · ~7 min

本頁8 分鐘閱讀
  1. 重點摘要
  2. 深入探討
  3. 查看測試時計算對一組問題的影響
  4. 現實世界的實施
  5. OpenAI o1 和 o3 推理模型的未來解釋
  6. 付諸實踐
  7. 資料來源與延伸閱讀
  8. 不斷探索
  9. 常見問題

重點摘要

  1. OpenAI o1於2024年9月發布,經過強化學習訓練,在回答之前產生隱藏的思維鏈。
  2. OpenAI 報告稱,在 AIME 2024 上,o1 一次嘗試的平均成績為 74%,而 GPT-4o 的成績為 12%。
  3. 推理模型透過更多的測試時間計算而得到改進,可以透過對每個答案思考更長的時間,或者透過對許多答案進行採樣並在其中進行選擇。
  4. o3 和 o4-mini 於 2025 年 4 月發布,可以在推理過程中使用網路搜尋、Python 和圖像工具。
  5. 隱藏推理令牌按輸出令牌計費,因此推理模型通常比聊天模型成本更高且回應更慢。
  6. 根據 TechCrunch 報導,OpenAI 的 PersonQA 測試發現 o3 比 o1 更容易產生幻覺,這表明較長的推理並不能保證準確性。

深入探討

OpenAI o1 and o3 are the first two generations of what OpenAI called its o-series, or reasoning models. Like earlier GPT models they are large language models: neural networks trained to predict the next token, a word or word fragment, of text. The difference is in what happens before the answer. A reasoning model first generates a long internal chain of thought, a running scratchpad of intermediate steps, checks and false starts, and only then writes its visible reply. OpenAI's announcement of o1 summarized the idea as a model that thinks before it answers. The thinking is produced as reasoning tokens, which OpenAI's API documentation says are not shown to developers but still take up room in the context window and are billed as output tokens.

The chain of thought is not just prompted; it is trained. OpenAI says o1 was trained with a large-scale reinforcement learning (RL) algorithm, a method in which a model tries many attempts and is rewarded for good outcomes rather than shown the correct text to copy. Through RL, according to the September 2024 announcement, the model learned to break hard steps into simpler ones, notice and correct its own mistakes, and switch strategies when an approach was failing. The company reported that o1's performance improved smoothly with more RL training compute and with more test-time compute, meaning more time spent thinking per question. That second axis was the headline: accuracy could be bought at inference time as well as by building a bigger model.

The timeline is short. OpenAI released o1-preview and o1-mini on 12 September 2024, followed by the full o1 in December 2024 alongside a system card describing its safety testing. On 20 December 2024 it previewed o3, and the ARC Prize Foundation published results for that preview on ARC-AGI-1, a puzzle benchmark designed to test novel pattern reasoning. o3-mini, a smaller and cheaper model aimed at science, math and coding, launched on 31 January 2025 and became the first reasoning model offered to free ChatGPT users. The full o3 and the smaller o4-mini arrived on 16 April 2025, and o3-pro followed for Pro subscribers in June 2025. OpenAI has since moved on: its posts now note that o1 is deprecated and that o3 and o4-mini are no longer available in ChatGPT and are deprecated in the API, with newer GPT-series reasoning models taking their place.

The published numbers explain the attention. On the 2024 AIME, a qualifying exam for the USA Mathematical Olympiad, OpenAI reported GPT-4o solved 12% of problems on average, while o1 averaged 74% with a single attempt, 83% with a majority vote over 64 samples, and 93% when re-ranking 1,000 samples with a learned scoring function. On Codeforces programming contests o1 reached the 89th percentile, and on GPQA Diamond, a set of graduate-level science questions, OpenAI said o1 was the first model to exceed the PhD experts it recruited for comparison. For o3, OpenAI reported 20% fewer major errors than o1 on difficult real-world tasks in external expert evaluations, and the ARC Prize Foundation reported the December 2024 preview scored 75.7% on its semi-private ARC-AGI-1 set at about $26 per task, and 87.5% using roughly 172 times more compute.

o3 also changed how reasoning connects to tools. OpenAI said o3 and o4-mini were trained with RL to decide when and how to use tools such as web search, a Python interpreter and image generation, and could chain several tool calls inside one reasoning process. They were also the first OpenAI models to put images into the chain of thought, cropping or rotating a picture mid-reasoning. With a Python interpreter, OpenAI reported o4-mini reached 99.5% on AIME 2025 on a single attempt, while cautioning that tool-assisted scores should not be compared with scores without tools.

Compared with a standard chat model, a reasoning model trades speed and cost for accuracy on problems with checkable answers. OpenAI's own human preference tests found o1-preview strongly preferred for data analysis, coding and math but not preferred on some natural language tasks such as writing. Compared with open alternatives, DeepSeek-R1, released in January 2025 and later published in Nature, showed that similar reasoning behavior, including self-reflection and verification, could emerge from reinforcement learning without human-written reasoning examples, and its authors described the training method publicly, whereas OpenAI has not published the o-series architecture or training recipe.

The limitations are well documented. The raw chain of thought is hidden: OpenAI chose to show users only a model-written summary, citing user experience, competitive advantage and the goal of monitoring the unaltered reasoning for problems such as manipulation. More thinking does not guarantee more truth. TechCrunch, citing OpenAI's o3 system card, reported that on OpenAI's PersonQA test of facts about people o3 hallucinated in 33% of answers and o4-mini in 48%, against 16% for o1, and the nonprofit lab Transluce found o3 sometimes claimed to have run code it could not have run. Benchmark context matters too: the ARC Prize Foundation noted that the o3 it tested had been trained on part of the ARC public training set and was not the same version later released.

In practice, reasoning models cost more per answer than chat models because every hidden reasoning token is billed. OpenAI's documentation recommends reserving at least 25,000 tokens for reasoning and output when starting out, warns that a response can run out of room before any visible text appears, and offers a reasoning effort setting and a cap on output tokens to control spending. Latency ranges from seconds to minutes. For simple lookups or casual writing, a faster non-reasoning model is usually the better fit.

03工作範例

查看測試時計算對一組問題的影響

  1. 選擇 15 個具有精確數字答案的問題,例如過去的 AIME 考試,這樣每個答案都可以自動檢查。

  2. 以較低的推理工作量向推理模型提出每個問題一次,並記錄 API 使用欄位中報告的準確性、延遲和推理令牌計數。

  3. 在高推理強度下重複並比較:準確性通常會提高,推理令牌、成本和延遲也會隨之增加。

  4. 對於並行方法,以中等努力對每個問題進行 8 次採樣,並將最常見的答案作為多數票。

  5. 透過將總代幣乘以公佈的價格並除以正確的數量,計算每個設定的每個正確答案的成本。

  6. 以非推理聊天模型作為基準重複整個測試。

它顯示了什麼

每個設定的準確度、延遲時間和每個正確答案的成本的小表。它應該顯示為 o1 報告的模式 OpenAI,準確性隨著測試時間計算而提高,並具體說明該增益的價格。

現實世界的實施

解決 AIME 2024 競賽數學問題:OpenAI 報告稱,o1 在一個樣本中平均解決了 74% 的問題,而 GPT-4o 為 12%,對 1,000 個樣本重新排名時為 93%。

競爭性程式設計:根據 o1 進行微調的模型在 2024 年國際資訊學奧林匹克競賽規則下獲得了 213 分,排名第 49 位。

ChatGPT 與 o3 的多步驟研究,OpenAI 表示可以搜尋網絡,對上傳的資料執行 Python 並產生圖表,將工具呼叫連結到一個答案。

使用 o3 和 o4-mini 閱讀模糊的白板照片或手繪圖表,OpenAI 表示可以裁剪、縮放和旋轉圖像作為推理的一部分。

使用 o3-mini 的三種推理工作設定(低、中、高)調整 API 中的成本和延遲,因此簡單的請求會簡短地思考,困難的請求會思考更長的時間。

OpenAI o1 和 o3 推理模型的未來解釋

OpenAI described o3 as pushing an additional order of magnitude of compute into both RL training and inference-time reasoning while still seeing gains, and it has since folded reasoning into newer GPT-series models rather than a separate o-series line. Several questions remain open. The hallucination results reported for o3 and o4-mini in April 2025 show that longer reasoning can raise confident errors on factual questions, and the cause is not settled. Chain-of-thought monitoring, which OpenAI cited as a reason to keep reasoning hidden but unaltered, depends on the chain staying faithful to the model's actual computation, which is an active research topic. Benchmarks are saturating: the ARC Prize Foundation called ARC-AGI-1 saturating in 2024 and launched the harder ARC-AGI-2, on which it said early o3 data pointed to under 30% even at high compute. Open-weight work that began with DeepSeek-R1 in January 2025 has made reasoning training reproducible outside OpenAI, so cost per correct answer, not only peak scores, is becoming the main point of comparison.

付諸實踐

為什麼這很重要

  • 供應商策略. 供應商路線圖會影響您的團隊接下來可以建立的功能。
  • 成本與預算. 商業條款和部署選項會影響長期成本和風險。
  • 風險與安全. 公司激勵措施塑造了產品預設、安全態勢和開放性。

留意

  • 發佈公告可能會超過實際生產工作流程的穩定性。
  • API 定價或政策轉變可能會在一夜之間打破假設。
  • 單一供應商依賴性增加了鎖定和遷移成本。

後續步驟

  1. 使用您自己的任務和資料集評估提供者。
  2. 在整合之前查看隱私、安全和法律條款。
  3. 維護跨模型或供應商的後備計劃。
  4. 監控發行說明,以便路線圖的變更不會讓團隊感到意外。

資料來源與延伸閱讀

  1. OpenAI向法學碩士學習推理
  2. OpenAI介紹 OpenAI o3 和 o4-mini
  3. OpenAIOpenAI o3-mini
  4. OpenAI API documentation推理模型
  5. arXiv (OpenAI)OpenAI o1系統卡
  6. ARC Prize FoundationOpenAI o3 ARC-AGI-Pub 突破高分
  7. TechCrunchOpenAI全新推理AI模型產生更多幻覺
  8. arXiv (published in Nature, 2025)DeepSeek-R1:透過強化學習激勵大型語言模型的推理能力

不斷探索

免費電子報

獲取每日 AI 簡報

每個工作日上午提供三則經過驗證的 AI 新聞,以淺顯易懂的英文撰寫。永久免費,無廣告。

每個工作日一封電子郵件。一鍵取消訂閱。我們絕不出售或分享您的地址。

測試一下自己

參加 OpenAI o1 和 o3 推理模型解釋測驗

每個答案皆有即時回饋,通過課程後可獲得具備可驗證 ID 的可分享證書。

開始測驗

支持免費的 AI 教育。AI Understanding 是一個 501(c)(3) 非營利組織——絕無廣告,絕無付費牆。 捐款

常見問題

OpenAI o1 和 o3 推理模型解釋是什麼?

OpenAI o1 和 o3 是推理模型,它們在回答之前使用額外的測試時計算來解決數學、科學和編碼中的多步驟問題。

OpenAI 的 o 系列文件說模型在返回最終答案之前使用什麼?

OpenAI 描述了內部推理令牌,但這並不意味著使用者收到完整的人類可讀的推理記錄。

OpenAI所說的o1系列訓練方法的一部分是什麼?

OpenAI 將強化學習描述為複雜推理 o1 訓練的一部分;避免將特定的訓練公開推廣到每個後續模型。

推理時間計算縮放對於推理模型意味著什麼?

OpenAI 描述了額外的推理計算,但測量的增益取決於任務,並且可能會增加令牌使用或延遲。

應如何解釋 OpenAI 推理模型報告的基準分數?

分數取決於模型版本、推理工作、基準設定和評估工具;不要將一個數字推廣到所有任務或版本。

為什麼即使使用者看不到內部推理令牌也會影響 API 成本?

OpenAI 文件解釋說,即使最終回應中未顯示,推理令牌也會消耗輸出令牌和上下文容量。