Arcs Directions
Back to Timeline
2024-09-12

OpenAI releases o1, the first reasoning model that 'thinks before it answers'

Capability Breakthrough

Event Summary

OpenAI o1 (September 2024) was the first reasoning model that ‘thought before answering.’ How inference-time compute improved AIME math scores from 13% to 83%.

Impact Assessment

  • Capability Leap +3 · Long-term

    Introduced test-time compute scaling as a new axis of AI progress. o1 achieved PhD-level accuracy on science benchmarks (GPQA Diamond) and solved 12/15 AIME math problems vs. GPT-4o's 1.8/15. The chain-of-thought reasoning approach generalized across chemistry, physics, biology, math, and coding — matching or exceeding expert human performance on hard reasoning tasks.

    Affected Groups: AI researchers, mathematicians, scientists, software engineers

  • Paradigm Shift +3 · Long-term

    Changed the AI field's understanding of scaling from a single axis (training compute) to two axes (training + inference compute). The concept of 'reasoning models' became a new category. Labs including Google, Anthropic, and DeepSeek all launched their own reasoning models in response. The idea that giving a model more time to 'think' improves performance had profound implications for AI system design, pricing, and capability forecasting.

    Affected Groups: entire AI field, researchers, AI companies, policymakers

  • Risk Creation -1 · Medium-term

    Hidden chain-of-thought reasoning raised transparency concerns — users could see the model's output but not its reasoning process. The model's improved resistance to jailbreaking was positive for safety, but the potential for deceptive reasoning (the model could 'think' harmful things without revealing them) created new AI governance challenges.

    Affected Groups: AI safety researchers, policymakers, ethicists, general public

Consensus & Sources

Significance L2
Category Capability Breakthrough
Consensus Broad Consensus
Impact Index 7/10
  • 1
    Introducing OpenAI o1 — OpenAI

    URL: https://openai.com/index/introducing-openai-o1-preview/

    We've developed a new series of AI models designed to spend more time thinking before they respond. They can reason through complex tasks and solve harder problems than previous models.
    Reference Evidence Citation logged Live source
  • 2
    OpenAI releases o1, its first model with reasoning abilities — The Verge

    URL: https://www.theverge.com/2024/9/12/24242439/openai-o1-model-reasoning-strawberry-chatgpt

    OpenAI releases o1, its first model with 'reasoning' abilities.
    News Report Citation logged Live source

Related Events

← Origins
2017-12 L3

"Attention Is All You Need" — the Transformer architecture is born

Eight Google researchers published 'Attention Is All You Need' at NeurIPS 2017, introducing the Transformer architecture. By dispensing with recurrence and convolution entirely and relying solely on self-attention mechanisms, the Transformer achieved state-of-the-art translation results while training significantly faster. The paper has been cited over 250,000 times, and the Transformer became the foundation of every major AI system since—from GPT and BERT to AlphaFold and modern speech models.

2023-03-14 L2

OpenAI releases GPT-4, the first multimodal large language model

On March 14, 2023, OpenAI released GPT-4 — a multimodal model that could accept both text and image inputs while generating text outputs. GPT-4 scored in the 90th percentile on the bar exam, the 99th percentile on the SAT, and demonstrated dramatically improved reasoning, factuality, and steerability over GPT-3.5. It was integrated into ChatGPT Plus as a premium tier, and its API became the default for developers building on LLMs. GPT-4 raised the bar on what LLMs could do and set a new baseline for AI capability that persisted until the model release wave of late 2024.

2024-05-13 L1

OpenAI releases GPT-4o, bringing real-time voice conversation to everyone

On May 13, 2024, OpenAI released GPT-4o ('omni'), a natively multimodal model combining text, vision, and audio processing in a single neural network. For the first time, users could have real-time voice conversations with an AI that could express emotion, laugh, pause, and adjust its tone — indistinguishable from human conversation. GPT-4o matched GPT-4 on text benchmarks while dramatically improving voice and vision performance. The model was made free for all ChatGPT users, effectively doubling the available AI capability overnight.

→ Successors
2024-12-11 L1

Google launches Gemini 2.0 with Project Mariner, unveiling a vision for AI agents

On December 11, 2024, Google DeepMind launched Gemini 2.0, its 'agentic era' AI model, alongside three agent projects: Project Mariner (browser agent), Project Astra (universal assistant prototype), and Jules (coding agent). The launch signalled Google's formal entry into the AI agent race, bringing agent-native capabilities to billions of users through Google's existing product ecosystem.

2025-01-20 L2

DeepSeek releases R1, an open-source reasoning model that shocks global markets

On January 20, 2025, Chinese AI lab DeepSeek released R1, an open-source reasoning model matching OpenAI's o1 on key benchmarks at a fraction of the training cost. Using Mixture-of-Experts architecture and Group Relative Policy Optimization, R1 achieved o1-level reasoning for approximately 1/20th of the training compute. On January 27, the release triggered a global stock market selloff that erased $593 billion from Nvidia's market cap — the largest single-day value loss in history — and DeepSeek became the most-downloaded free app on the US iOS App Store.

2025-01-23 L1

OpenAI launches Operator, the first AI agent that can browse the web autonomously

On January 23, 2025, OpenAI released Operator, a research preview of an AI agent that could autonomously navigate a web browser to perform tasks — ordering groceries, making dinner reservations, filling out forms. Powered by a Computer-Using Agent (CUA) model combining GPT-4o vision with o-series reasoning, Operator marked OpenAI's entry into the agent space. By July 2025, it was fully integrated into ChatGPT, making autonomous browsing a standard feature for hundreds of millions of users.

2025-07 L2

Google DeepMind's Gemini Deep Think achieves gold medal at the International Mathematical Olympiad

In July 2025, Google DeepMind announced that an advanced version of Gemini with Deep Think had achieved a gold-medal standard at the International Mathematical Olympiad, scoring 35 out of 42 points. This marked the first time an AI system reached the highest level of the world's most prestigious mathematics competition for pre-university students — solving problems requiring deep mathematical reasoning, creativity, and multi-step logic. The achievement demonstrated that inference-time compute scaling could match the most challenging human intellectual benchmarks.

2025-08-07 L2

OpenAI releases GPT-5, unifying reasoning, multimodality, and task execution

On August 7, 2025, OpenAI released GPT-5, a unified AI system that combined advanced reasoning, native multimodality, and autonomous task execution — eliminating the need for users to switch between separate models (GPT-4o for chat, o-series for reasoning, DALL-E for images). According to Sam Altman, GPT-5 could 'provide PhD-level expertise' across domains. The release marked the culmination of OpenAI's journey from a single-purpose language model (GPT-3, 2020) to a general-purpose AI system capable of reasoning, creating, and acting.

Local Lineage OpenAI releases o1, ... 2024-09-12 · L2 "Attention Is All ... 2017-12 · L3 OpenAI releases GP... 2023-03-14 · L2 OpenAI releases GP... 2024-05-13 · L1 Google launches Ge... 2024-12-11 · L1 DeepSeek releases ... 2025-01-20 · L2 OpenAI launches Op... 2025-01-23 · L1 Google DeepMind's ... 2025-07 · L2