GenAI Chronicle
2010
2011
2013
2014
2015
2016
2017
2018
2019
2020
2021
2022
2023
2024
2025
2026
2024

December

7 events

Editors' summary

OpenAI ships for twelve days straight, Google bets on agents

On 5 December, OpenAI began "12 Days of OpenAI." Day one brought ChatGPT Pro at $200 a month; Sora went public on the 9th, and on the 20th the company showed o3, its next reasoning model, reporting a large jump on evaluations built from hard problems.

On the 11th, Google announced Gemini 2.0 under the banner of an "agentic era," showing prototypes that watch the screen and act on the user's behalf.

On the 27th, OpenAI published its plan to reorganise as a public benefit corporation, restructuring the arrangement in which a non-profit controls the commercial subsidiary.

On the 26th, China's DeepSeek released V3. How little it had reportedly cost to train drew attention.

On the BG2 podcast that month, Nadella said business applications would "collapse" in the agent era. The words were never his, but the remark took on a life of its own as "SaaS is dead."

This block is written by the editors. It is kept separate from the sourced record below.

Record7 events
  1. –
    CultureMicrosoft / Satya Nadella

    Nadella's remark spawns "SaaS is dead"

    On the BG2 podcast, Microsoft CEO Satya Nadella said that business applications are essentially "CRUD databases with a bunch of business logic," and that the category would collapse in an era where agents carry the logic. He never used the word "dead," but the remark took on a life of its own under the headline "SaaS is dead." Nadella repeated the argument in May 2025. The English phrase was adopted as-is in Japan, where "SaaS is Dead" became a stock question that SaaS companies kept being asked to answer.

  2. 5
    ProductOpenAI / Sam Altman

    ChatGPT Pro launches as "12 Days of OpenAI" begins

    OpenAI kicked off "12 Days of OpenAI," an event with announcements across 12 weekdays, launching on day one the $200-per-month ChatGPT Pro plan and the full version of its o1 reasoning model. Pro includes unlimited access to OpenAI o1, GPT-4o, and Advanced Voice, plus o1 pro mode, which uses more compute for the hardest problems. The event went on to deliver major releases including Sora's public launch and the o3 announcement.

  3. 9
    ProductOpenAI / Sam Altman

    Sora released to the public

    OpenAI released its video generation model Sora to the public, about 10 months after the February reveal. A faster version, Sora Turbo, became available at the dedicated site sora.com for ChatGPT Plus and Pro subscribers. It launched in the U.S. and most countries, though not initially in the EU and the UK. The release came on day three of the "12 Days of OpenAI" event.

  4. 11
    ModelGoogle / Sundar Pichai / Demis Hassabis

    Gemini 2.0 arrives, framed for "the agentic era"

    Google announced Gemini 2.0, positioning it as a model for "the agentic era," and rolled out an experimental Gemini 2.0 Flash globally across selected products. Multimodal, it could generate images and speak in a range of voices, with users able to ask it to talk faster or slower, and Google said it ran twice as fast as Gemini 1.5 Pro while performing better. Alongside came Project Mariner, a research prototype that reads what is on a browser screen — pixels, text, code, images, forms — and completes tasks through an experimental Chrome extension, scoring a then-best 83.5% as a single agent on WebVoyager, a benchmark of real-world web tasks. Other agent prototypes including the coding tool Jules were shown at the same time, making clear the industry's shift from AI that answers to AI that acts.

  5. 20
    ModelOpenAI / Sam Altman

    OpenAI announces next-generation reasoning model o3

    On the final day of "12 Days of OpenAI," OpenAI announced o3, the successor to its o1 reasoning model, along with the smaller o3-mini, opening applications for external safety researchers to test them ahead of general availability. OpenAI reported that o3 dramatically outscored earlier models on the ARC-AGI benchmark of generalization, underscoring the rapid pace of progress in reasoning models. The name "o2" was reportedly skipped for trademark reasons involving the British telecom O2.

  6. 26
    ModelDeepSeek

    DeepSeek-V3 arrives, trained for about $5.6 million

    China's DeepSeek released DeepSeek-V3, a sparse mixture-of-experts model with 671 billion parameters in total and 37 billion active per token, trained on 14.8 trillion tokens. The run took 2.788 million H800 GPU-hours, which the company put at roughly $5.6 million against contemporary estimates of $50–100 million to train GPT-4 — an order of magnitude less for performance close to the frontier. The weights were published. It was read as a lab cut off from the best GPUs by export controls squeezing efficiency out of the hardware it could get, which raised its own questions about how well the controls were working. Observers noted at the time that the $5.6 million covered the final training run only, not the research, failed attempts, or the cluster itself.

  7. 27
    GovernanceOpenAI / Sam Altman / Elon Musk

    OpenAI announces plan to restructure as a public benefit corporation

    OpenAI announced a plan to convert its for-profit arm into a Delaware public benefit corporation. Saying it needed "more capital than we'd imagined," the company framed the move as a structure better suited to raising funds. The nonprofit would continue as a large shareholder of the PBC and pursue charitable initiatives. The plan marked a major shift from the capped-profit structure adopted in 2019 and stirred debate, coming amid litigation brought by Elon Musk.

↑ 2024