> For the complete documentation index, see [llms.txt](https://yeasy.gitbook.io/llm_internals/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://yeasy.gitbook.io/llm_internals/di-si-bu-fen-mo-xing-yu-qian-yan-pian/13_decoder_models/summary.md).

# 本章小结

**GPT 系列**展示了规模扩展的力量——从 1.17 亿到未公开规模的 GPT-4，涌现能力的发现推动了行业向更大模型的竞赛。GPT-4o mini 改变了 API 定价格局，而 **o1 系列**开辟了推理时计算扩展的新维度，证明投入更多推理计算同样能大幅提升模型能力。

**Llama** 通过 RoPE、RMSNorm、SwiGLU、GQA 等架构改进确立了现代 LLM 的标准配方。Llama 3.1 的 405B 模型首次让开放权重模型追近闭源前沿，3.2 将 LLM 带到手机端侧，3.3 则以 70B 实现了 405B 级别的对齐效果，展现了开放生态的蓬勃发展。

**DeepSeek** 以创新的 MoE 架构和 FP8 训练大幅降低训练运行成本并达到前沿水平。特别是 **DeepSeek-R1-Zero** 展示了纯强化学习激发推理行为的可能性，正式 **DeepSeek-R1** 则采用 cold-start 数据和多阶段训练，并将推理模式蒸馏到小模型中，为开放推理模型开辟了新路径。**Gemini** 代表了原生多模态、百万级上下文和智能体化工作流的发展方向。**Claude** 以 Artifacts 和 Computer Use 推动了 AI 交互方式的创新，Opus 4.8（2026 年 5 月）延续 Opus 4.7 的软件工程、视觉理解和复杂多步任务能力，并刷新当前 Opus 代际；具体基准应以官方模型页和系统卡为准。**Qwen 2.5** 标志着中国开源 LLM 在国际基准上的全面崛起。

**T5/BART** 等编码器-解码器模型在 Seq2Seq 任务上有架构优势，但在大规模实践中让位于更简洁的纯解码器架构。

下一章将讨论 Transformer 架构的前沿创新与未来趋势。

***

> 📝 **发现错误或有改进建议？** 欢迎提交 [Issue](https://github.com/yeasy/llm_internals/issues) 或 [PR](https://github.com/yeasy/llm_internals/pulls)。


---

# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.

## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question.

Perform an HTTP GET request on the current page URL with the `ask` query parameter, and the optional `goal` query parameter:

```
GET https://yeasy.gitbook.io/llm_internals/di-si-bu-fen-mo-xing-yu-qian-yan-pian/13_decoder_models/summary.md?ask=<question>&goal=<endgoal>
```

`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is optional and describes the broader end goal you are ultimately trying to accomplish on behalf of the user. GitBook uses it to tailor the answer towards what is most useful for that goal.

The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.