steve sparks
17 of 81
ContentsPart II · Chapter 4

Chapter 4From early models to today

Thinking and doing, 2024 on

The newest stretch isn’t about writing better. It’s about doing things. It uses parts of the picture that later chapters explain, so here is a short preview. A tool is an action that the software around a model can take when the model asks, such as searching the web or opening a file (Chapter 8). The harness is that software (Chapter 7). An agent is a model working in a loop with tools to carry out a task, the pattern ReAct showed in 2022.1 For now, just notice when these arrived.

The entries below name products, as of October 2026. Names like these date quickly; they are here as examples of the trend, not as a guide to what to use.

  • September 2024. Reasoning models: models that work a problem through before answering. OpenAI said its o1 model “thinks before it answers.”2
  • November 2024. Anthropic introduced the Model Context Protocol (MCP), “a new standard for connecting AI assistants to the systems where data lives.”3
  • February and April 2025. Coding agents that edit code and run commands: Anthropic’s Claude Code in February and OpenAI’s open-source Codex CLI in April.456
  • August 2025. OpenAI released open-weight models that anyone could download and run.7
  • October 2025. Skills: in Anthropic’s words, “organized folders of instructions, scripts, and resources” that an agent loads when needed; they were later published as an open standard.8
  • April to June 2026. The most capable models reached different users with different safeguards. Anthropic first gave a preview only to selected security defenders, then in June released Fable 5 for everyone and Mythos 5 for vetted partners.910
  • June and July 2026. A U.S. export-control directive led Anthropic to pause its two newest models for all users on June 12. Fable 5 returned on July 1.1112
  • September 2026. Anthropic, OpenAI and Google each released new top models within a few weeks, among them Fable 5.1, GPT-6 Astra (at first to a limited set of organizations) and Gemini 4 Argon (at first only to trusted cyber defenders).131415
  • October 2026. Small models kept improving. Anthropic called Claude Haiku 5.5 “the cheapest, fastest, and most capable small model we’ve ever released.”16

Looking back from October 2026, a few things had changed. Models were sold for long-running agentic work on complex tasks.17 Researchers argued for teams of models: small ones for narrow, repetitive sub-tasks, larger ones where broad ability was needed.18 The most capable models, often called frontier models, were released with different safeguards for different users. Anthropic described Fable 5.1 and Mythos 5.1 as “the same model, but with different levels of safeguards.”19

Open standards moved to a neutral home. MCP went to the new Agentic AI Foundation, under the Linux Foundation, in December 2025, and Agent Skills followed in October 2026.2021 And a government took a direct role in who could use frontier models.11

Notes (21)
  1. 1Shunyu Yao, Jeffrey Zhao, Dian Yu, Nan Du, Izhak Shafran, Karthik Narasimhan, and Yuan Cao, “ReAct: Synergizing Reasoning and Acting in Language Models” (preprint, arXiv, October 6, 2022), https://doi.org/10.48550/arXiv.2210.03629.
  2. 2OpenAI, “Learning to reason with LLMs,” September 12, 2024, https://openai.com/index/learning-to-reason-with-llms/ (accessed October 27, 2026).
  3. 3Anthropic, “Introducing the Model Context Protocol,” November 25, 2024, https://www.anthropic.com/news/model-context-protocol (accessed October 27, 2026).
  4. 4Anthropic, “Claude 3.7 Sonnet and Claude Code,” February 24, 2025, https://anthropic.com/news/claude-3-7-sonnet (accessed October 27, 2026).
  5. 5OpenAI, “Introducing OpenAI o3 and o4-mini,” April 16, 2025, https://openai.com/index/introducing-o3-and-o4-mini/ (accessed October 26, 2026).
  6. 6OpenAI, “openai/codex,” GitHub, https://github.com/openai/codex (accessed October 26, 2026).
  7. 7OpenAI, “Introducing gpt-oss,” August 5, 2025, https://openai.com/index/introducing-gpt-oss/ (accessed October 27, 2026).
  8. 8Anthropic, “Equipping agents for the real world with Agent Skills,” Anthropic Engineering, October 16, 2025, https://www.anthropic.com/engineering/equipping-agents-for-the-real-world-with-agent-skills (accessed October 27, 2026).
  9. 9Anthropic, “Project Glasswing: Securing critical software for the AI era,” April 7, 2026, https://www.anthropic.com/glasswing (accessed October 27, 2026).
  10. 10Anthropic, “Claude Fable 5 and Claude Mythos 5,” June 9, 2026, https://www.anthropic.com/news/claude-fable-5-mythos-5 (accessed October 27, 2026).
  11. 11Anthropic, “Statement on the US government directive to suspend access to Fable 5 and Mythos 5,” June 12, 2026, https://www.anthropic.com/news/fable-mythos-access (accessed October 27, 2026).
  12. 12Anthropic, “Redeploying Claude Fable 5,” June 30, 2026, https://www.anthropic.com/news/redeploying-fable-5 (accessed October 27, 2026).
  13. 13Anthropic, “Release notes — Claude Platform,” Claude Platform Docs, https://platform.claude.com/docs/release-notes/overview (accessed October 26, 2026).
  14. 14OpenAI, “GPT-6 Astra: A new generation of intelligence,” September 2026, https://openai.com/index/gpt-6-astra/ (accessed October 27, 2026).
  15. 15K. Kavukcuoglu, “Gemini 4 Argon: our next era of frontier intelligence,” Google, September 30, 2026, https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-4-argon/ (accessed October 27, 2026).
  16. 16Anthropic, “Introducing Claude Haiku 5.5,” October 7, 2026, https://www.anthropic.com/claude-haiku-5-5 (accessed October 27, 2026).
  17. 17Anthropic, “Choosing the right model,” Claude Platform Docs, https://platform.claude.com/docs/en/about-claude/models/choosing-a-model (accessed October 26, 2026).
  18. 18Peter Belcak, Greg Heinrich, Shizhe Diao, Yonggan Fu, Xin Dong, Saurav Muralidharan, Yingyan Celine Lin, and Pavlo Molchanov, “Small Language Models are the Future of Agentic AI” (preprint, arXiv, June 2, 2025), https://doi.org/10.48550/arXiv.2506.02153.
  19. 19Anthropic, “Introducing Claude Fable 5.1 and Claude Mythos 5.1,” September 2026, https://www.anthropic.com/claude-fable-and-mythos-5-1 (accessed October 27, 2026).
  20. 20Anthropic, “Donating the Model Context Protocol and establishing the Agentic AI Foundation,” December 9, 2025, https://www.anthropic.com/news/donating-the-model-context-protocol-and-establishing-of-the-agentic-ai-foundation (accessed October 27, 2026).
  21. 21Anthropic, “Donating Agent Skills to the Agentic AI Foundation,” Anthropic News, October 22, 2026, https://www.anthropic.com/news/donating-agent-skills-to-the-agentic-ai-foundation (accessed October 26, 2026).

Chapter 4 · page 4 of 5

Turn pages with the ← and → keys.

Cite this page

https://aifromthepartsup.com/from-early-models-to-today/4/

Using this book

© 2026 Steve Sparks. All rights reserved. Teachers and students may assign, read, quote and excerpt chapters of this book for classes and study, with credit to Steve Sparks and a link to aifromthepartsup.com. Anyone may quote short passages and cite any page, with a link back. To repost, reproduce or sell the book or larger parts of it, ask through GitHub.

Quoted material and figures from other sources remain their owners'.