AI Model Timeline

Sources and attribution

The timeline cites 230 sources. 14 of them have terms that permit reuse with credit. For the other 216, reuse terms are unknown, cite-only or restrictive, so we link to them and record individual facts (a date, a price, a score) with a short quotation of at most 200 characters, and copy nothing else. Where a source's terms prohibit even a quotation, none is shown.

Each value on the timeline links to its source in the evidence drawer. The credits below are generated from the same reviewed dataset, so they cannot drift from it. Where a credit says third-party data keeps its original licence, that licence still applies.

Credits

• Aakanksha Chowdhery et al., Google Research, PaLM: Scaling Language Modeling with Pathways, arXiv:2204.02311v5, CC BY 4.0. Individual findings summarized; no table reproduced.

  • PaLM: Scaling Language Modeling with Pathways

• Aider / Paul Gauthier and contributors, dedicated original o1 polyglot results and accompanying launch article, pinned at repository commit 5dc9490bb35f9729ef2c95d00a19ccd30c26339c. Apache License 2.0. Nettacular extracted one reported result and normalized its fields. Preserve applicable source copyright, patent, trademark, attribution and NOTICE notices, provide the license with redistributed source material, and identify modifications.

  • Aider original o1 polyglot leaderboard (pinned YAML)

• ARC Prize Foundation. ARC-AGI-2. Apache License 2.0; retain applicable copyright, license and attribution notices when redistributing licensed source material.

  • Abstraction and Reasoning Corpus for Artificial General Intelligence v2 (ARC-AGI-2)

• Credit Epoch AI and source authors; third-party data keeps original licensing.

  • Claude Haiku 4.5
  • Claude Haiku 5.5

• Credit Epoch AI and the source authors. Externally sourced data retains its original licensing.

  • About our benchmarking

• Dan Hendrycks et al., Measuring Massive Multitask Language Understanding, arXiv:2009.03300v3. Citation only; arXiv distribution license does not establish downstream reuse permission.

  • Measuring Massive Multitask Language Understanding

• Epoch AI

  • Claude 3.7 Sonnet
  • Claude Fable 5 | Epoch AI
  • Claude Sonnet 4
  • Claude Sonnet 4.5
  • Claude Sonnet 5
  • Claude Sonnet 5.5 | Epoch AI

• Francois Chollet. ARC-AGI-1. Apache License 2.0; retain applicable copyright, license and attribution notices when redistributing licensed source material.

  • Abstraction and Reasoning Corpus for Artificial General Intelligence v1 (ARC-AGI-1)

• Gemini Team, Google, Gemini 2.5 technical report, arXiv:2507.06261v6, CC BY 4.0. Individual findings summarized; no table reproduced.

  • Gemini 2.5: Pushing the Frontier with Advanced Reasoning, Multimodality, Long Context, and Next Generation Agentic Capabilities

• Gemini Team, Google, Gemini: A Family of Highly Capable Multimodal Models, arXiv:2312.11805v5, CC BY 4.0. This draft extracts individual reporting facts without reproducing the table.

  • Gemini: A Family of Highly Capable Multimodal Models

• Hyung Won Chung et al., Scaling Instruction-Finetuned Language Models, arXiv:2210.11416v5, CC BY 4.0. Changes: selected published results transcribed into structured observations.

  • Scaling Instruction-Finetuned Language Models

• Inspect Evals. Repository license: Copyright (c) 2024 UK AI Security Institute, MIT License. Preserve copyright and permission notices when redistributing substantial licensed material.

  • MMLU: Measuring Massive Multitask Language Understanding – Inspect Evals

• Jia He et al., Does Prompt Formatting Have Any Impact on LLM Performance?, arXiv:2411.10541v1, CC BY 4.0.

  • Does Prompt Formatting Have Any Impact on LLM Performance?

• Mark Chen et al., OpenAI, Evaluating Large Language Models Trained on Code, arXiv:2107.03374v2.

  • Evaluating Large Language Models Trained on Code

• MMMU Team

  • MMMU

• OpenAI

  • Function calling and other API updates
  • GPT-2: 1.5B release
  • GPT-2: 6-month follow-up
  • GPT-3.5 Turbo Model
  • GPT-4
  • GPT-4 API general availability and deprecation of older models in the Completions API
  • GPT-4o Mini Model
  • GPT-4o mini: advancing cost-efficient intelligence
  • Hello GPT-4o
  • Introducing ChatGPT
  • Introducing ChatGPT and Whisper APIs
  • New models and developer products announced at DevDay
  • OpenAI API
  • OpenAI robots.txt
  • OpenAI’s API now available with no waitlist
  • Terms of Use

• princeton-nlp, SWE-bench_Verified dataset card and repository revision c104f840cc67f8b6eec6f759ebc8b2693d585d4a.

  • princeton-nlp/SWE-bench_Verified

• Shen Zheng, Yuyu Zhang, Yijie Zhu, Chenguang Xi, Pengyang Gao, Xun Zhou, and Kevin Chen-Chuan Chang. GPT-Fathom: Benchmarking Large Language Models to Decipher the Evolutionary Path towards GPT-4 and Beyond. ACL Anthology 2024.findings-naacl.87. Creative Commons Attribution 4.0 International License.

  • GPT-Fathom: Benchmarking Large Language Models to Decipher the Evolutionary Path towards GPT-4 and Beyond

• Sirui Hong et al., MetaGPT: Meta Programming for a Multi-Agent Collaborative Framework, arXiv:2308.00352v6, CC BY 4.0.

  • MetaGPT: Meta Programming for a Multi-Agent Collaborative Framework

• Surge AI, arXiv:2608.10677v1, CC BY 4.0.

  • Chartography: A Benchmark for Professional Chart Understanding

• Vectara, Hallucination Leaderboard, README.md at commit 5107e1e7299bf8529c83cbd30b4355463a6fcab5, Apache License 2.0. Retain applicable copyright, attribution and NOTICE notices, include the license, and identify adaptations.

  • Hallucination Leaderboard

• Xiang Yue et al., MMMU, arXiv:2311.16502v1. Individually cited facts only.

  • MMMU: A Massive Multi-discipline Multimodal Understanding and Reasoning Benchmark for Expert AGI

• Xue Jiang, Yihong Dong, Lecheng Wang, Zheng Fang, Qiwei Shang, Ge Li, Zhi Jin, and Wenpin Jiao. Self-planning Code Generation with Large Language Models. arXiv:2303.06689v5. The linked arXiv non-exclusive distribution license does not establish a public reuse grant; cite-only treatment.

  • Self-planning Code Generation with Large Language Models

• Ziqian Bi et al., Is GPT-OSS Good? A Comprehensive Evaluation of OpenAI's Latest Open Source Models, arXiv:2508.12461v3, CC BY 4.0. Structured factual extraction; no endorsement implied.

  • Is GPT-OSS Good? A Comprehensive Evaluation of OpenAI's Latest Open Source Models