Back to News

Zhipu discloses GLM-6.0 'fully self-training' plan in HKEX filing

#zhipu#glm-6.0#recursive-self-improvement#funding

In a Hong Kong Stock Exchange filing, Zhipu disclosed plans to invest about HK$23.5 billion (60% of net proceeds from a HK$39.3 billion fundraising) into next-generation GLM foundation models and its 'Fully Self Training' system, a recursive self-improvement loop covering data self-generation, environment self-creation, and infrastructure self-optimization. The filing also previewed GLM-6.0's technical direction, including scaling effective model size, native multimodal unified modeling, and deeper effective computation without excessive inference cost increases.

Coverage timeline

  1. 量子位量子位
    智谱提前剧透GLM-6.0：完全自训练方法公开了

    智谱提前剧透GLM-6.0：完全自训练方法公开了 梦晨 2026-09-14 08:30:56 来源： 量子位 模型、论文都还没出，没想到先从财务公告上剧透了 梦晨 发自 凹非寺 量子位 | 公众号 QbitAI 上市不到九个月，智谱一口气再融 335亿元 。 智谱在港交所发布公告：配售新H股加发行201.4亿元人民币零息可转债，合计预计净募 约393亿港元 。 6成资金指向同一个目标， 递归式自我改进（Recursive Self-Improvement，RSI） ， 但在智谱这里有一个新名字， “完全自训练” 。 在8月底的半年报业绩沟通会上，唐杰就曾回应“只会后训练”质疑，透露过下一代GLM-6.0瞄准自进化。 现在模型、论文都还没出，没想到先从财务公告上剧透了。 智谱的完全自训练 300多亿元要怎么花，公告给出了明确的分配方案。 约 60% 的所得款项净额（约235亿港元）将投入下一代GLM基础模型和“完全自训练”体系的研发，以及大规模训练、推理和算力基础设施的部署与升级。 “完全自训练”（Fully Self Training）是智谱对RSI的具体技术表述，公告释义部分给出了定义： 下一代GLM模型将在上一代GLM所搭建的环境里训练，形成递归式自我改进（Recursive Self-Improvement）闭环，涵盖 数据自产、环境自造、基础设施自我优化 三个维度。 数据自产，不再完全依赖人类标注。 通过模型间自我对弈、规则校验、执行验证、模型评审和人工抽检等方式，自动生成、筛选和积累高质量训练数据，持续回流到预训练、中训练和后训练各阶段。 环境自造，让 智能体 去采集和转化真实世界的任务。 智能体自己试做、自己生成验证器、自己检查任务是否可解，不断扩充训练环境的数量、类型和复杂度。要让任务环境成为“可规模化生产及复用的训练资源”。 基础设施自我优化，模型已经会写代码了，那就让它帮忙优化自己赖以运行的训练和推理系统。 利用模型在编码和系统工程方面的能力，协助开发和优化训练及推理系统的算子、内核、调度、缓存和服务栈，使模型能够逐步参与支撑其自身迭代的基础设施升级。 闭环之外，模型本身的能力方向也在公告中预告。 智谱下一步要做的事情包括提升基础模型的有效规模和原生多模态统一建模能力，在不过度增加推理成本的前提下加深有效计算深度，让模型能跑更长链条的推理、规划和

Related stories

  • Willison keynote: Claude Opus 4.5 and GPT-5.1 mark 2026 LLM trends

    Simon Willison delivered the closing keynote at WeAreDevelopers World Congress North America in San Jose, summarizing key LLM trends and releases in 2026 so far. He highlighted the November 2025 releases of Claude Opus 4.5 and GPT-5.1, which enabled coding agents like Claude Code and Codex to cross a threshold of practical usability. The talk is available on YouTube with annotated slides and notes.

    1 reports · 1 sources

  • Z.ai debuts GLM-5.3 with post-training scaling gains, open weights in two weeks

    Z.ai (Zhipu AI) announced GLM-5.3, an open-source model built on the same base as GLM-5.2 with substantially extended post-training, yielding a 50% gain on internal coding benchmarks and state-of-the-art results among open models on Terminal Bench 3.0 and Agents' Last Exam (CLI). The model matches Mythos 5 in white-box code review and vulnerability discovery, and scored 60 on the Artificial Analysis Intelligence Index, tying Kimi K3. Model weights will be released in two weeks after security hardening; API pricing is set at $1.40 per million input tokens and $4.40 per million output tokens.

    9 reports · 6 sources1154 points · 573 comments

  • Z.ai confirms Ox Alpha is GLM-5.3 Flash, releases open weights

    Z.ai confirmed that the anonymous model 'Ox Alpha', which went viral on OpenRouter, is a new iteration of its GLM series, specifically GLM-5.3 Flash, and released its weights on Hugging Face under a new license. The model, a 320B-parameter natively multimodal model running on domestic chips, topped OpenRouter's leaderboard and ended DeepSeek's 56-day streak on OpenCode. The license requires companies with over $10B in revenue to pass a security review.

    9 reports · 4 sources801 points · 281 comments

  • Zhipu AI open-sources GLM-5.3-Flash, a 320B multimodal model priced at 1/40 of Claude Opus 4.8

    Zhipu AI released and open-sourced GLM-5.3-Flash, the first natively multimodal model in the GLM-5 series, with 320B total parameters (18B active). It scores 57 on the Artificial Analysis Intelligence Index, matching Claude Opus 4.8, and is priced at one-tenth of GLM-5.3, or one-fortieth of Opus 4.8. The model uses a hybrid sparse and linear attention architecture, runs on domestic Chinese chips, and was tested anonymously as Ox-Alpha on OpenRouter and OpenCode.

    8 reports · 5 sources1126 points · 573 comments

News updates every hour. Subscribe in your RSS reader to get new stories automatically.

Subscribe