noscript main { max-width: 1180px; margin: 0 auto; padding: 32px 24px; color: #6b7280; font: 16px/1.6 system-ui, -apple-system, BlinkMacSystemFont, "Segoe UI", sans-serif; } noscript h1 { margin: 0 0 28px; color: #111827; font-size: 28px; line-height: 1.25; } noscript a { color: #0284c7; text-decoration: underline; text-underline-offset: 2px; } noscript .gf-crawler-header { border-bottom: 1px solid #e5e7eb; } noscript .gf-crawler-header-inner { max-width: 1180px; margin: 0 auto; padding: 18px 24px; } noscript .gf-crawler-home-link { color: #0284c7; font-size: 18px; font-weight: 600; } noscript .gf-crawler-topic-list { width: 100%; } noscript .gf-crawler-topic-list-head, noscript .gf-crawler-topic-row { display: grid; grid-template-columns: minmax(0, 1fr) 120px 120px 160px; } noscript .gf-crawler-topic-list-head { border-bottom: 2px solid #e5e7eb; color: #9ca3af; font-size: 15px; font-weight: 600; } noscript .gf-crawler-topic-list-head span { padding: 0 12px 10px; } noscript .gf-crawler-topic-list-head span:not(:first-child) { text-align: center; white-space: nowrap; } noscript .gf-crawler-topic-items { margin: 0; padding: 0; list-style: none; } noscript .gf-crawler-topic-row { border-bottom: 1px solid #e5e7eb; } noscript .gf-crawler-topic-row > div { padding: 20px 12px; align-self: center; } noscript .gf-crawler-visually-hidden { position: absolute; width: 1px; height: 1px; padding: 0; margin: -1px; overflow: hidden; clip: rect(0, 0, 0, 0); white-space: nowrap; border: 0; } noscript .gf-crawler-topic-title { margin: 0 0 4px; font-size: 20px; line-height: 1.35; } noscript .gf-crawler-topic-meta { margin: 0 0 12px; color: #6b7280; font-size: 14px; } noscript .gf-crawler-topic-meta a { margin-right: 8px; } noscript .gf-crawler-topic-excerpt { max-width: 760px; margin: 0; color: #8b9098; } noscript .gf-crawler-topic-number, noscript .gf-crawler-topic-activity { color: #8b9098; text-align: center; white-space: nowrap; } noscript .gf-crawler-topic-page { max-width: 920px; } noscript .gf-crawler-topic-head { padding-bottom: 24px; border-bottom: 1px solid #e5e7eb; } noscript .gf-crawler-topic-summary { max-width: 760px; margin: -14px 0 18px; color: #8b9098; font-size: 17px; } noscript .gf-crawler-post-meta { margin: 0 0 18px; color: #8b9098; font-size: 14px; } noscript .gf-crawler-post-meta a, noscript .gf-crawler-post-meta span, noscript .gf-crawler-post-meta time { margin-right: 12px; } noscript .gf-crawler-post-body { max-width: 760px; color: #374151; } noscript .gf-crawler-post-body img, noscript .gf-crawler-post-body video { display: block; max-width: 100%; max-height: 560px; width: auto; height: auto; object-fit: contain; } noscript .gf-crawler-post-body figure { margin: 16px 0; max-width: 100%; } noscript .gf-crawler-post-body table { display: block; max-width: 100%; overflow-x: auto; border-collapse: collapse; } noscript .gf-crawler-post-body pre { overflow: auto; padding: 12px; background: #f3f4f6; } noscript .gf-crawler-post-list { margin-top: 28px; } noscript .gf-crawler-post-list h2 { margin: 0 0 8px; color: #111827; font-size: 20px; } noscript .gf-crawler-post-list ol { margin: 0; padding: 0; list-style: none; } noscript .gf-crawler-post { padding: 18px 0; border-top: 1px solid #e5e7eb; } noscript .gf-crawler-user-page { max-width: 980px; } noscript .gf-crawler-user-head { display: flex; gap: 18px; align-items: center; padding-bottom: 22px; border-bottom: 1px solid #e5e7eb; } noscript .gf-crawler-user-avatar { width: 88px; height: 88px; border-radius: 50%; object-fit: cover; background: #f3f4f6; } noscript .gf-crawler-user-head h1 { margin: 0 0 4px; } noscript .gf-crawler-user-meta, noscript .gf-crawler-user-bio { margin: 0; color: #8b9098; } noscript .gf-crawler-user-bio { margin-top: 10px; max-width: 680px; color: #6b7280; } noscript .gf-crawler-user-stats dl { display: grid; grid-template-columns: repeat(4, minmax(0, 1fr)); gap: 12px; margin: 20px 0 32px; } noscript .gf-crawler-user-stats div { padding: 12px 0; border-bottom: 1px solid #e5e7eb; } noscript .gf-crawler-user-stats dt { color: #8b9098; font-size: 13px; } noscript .gf-crawler-user-stats dd { margin: 2px 0 0; color: #374151; font-size: 20px; } noscript .gf-crawler-user-section { margin-top: 28px; } noscript .gf-crawler-user-section h2 { margin: 0 0 12px; color: #111827; font-size: 20px; } noscript .gf-crawler-activity-list { margin: 0; padding: 0; list-style: none; } noscript .gf-crawler-activity { padding: 14px 0; border-top: 1px solid #e5e7eb; } noscript .gf-crawler-activity time { display: block; margin-top: 4px; color: #8b9098; font-size: 14px; } noscript .gf-crawler-error { max-width: 560px; min-height: 224px; margin: 0 auto; display: flex; flex-direction: column; align-items: center; justify-content: center; text-align: center; padding: 48px 24px; } noscript .gf-crawler-error-code { width: 32px; height: 32px; display: flex; align-items: center; justify-content: center; border-radius: 999px; background: #f3f4f6; color: #8b9098; font-size: 12px; font-weight: 600; letter-spacing: 0.05em; } noscript .gf-crawler-error h1 { margin: 8px 0 4px; color: #111827; font-size: 20px; line-height: 1.4; } noscript .gf-crawler-error-message { margin: 0; color: #8b9098; font-size: 14px; } noscript .gf-crawler-error-actions { margin: 16px 0 0; } noscript .gf-crawler-footer { max-width: 1180px; margin: 24px auto 0; padding: 20px 24px 32px; border-top: 1px solid #e5e7eb; color: #374151; } noscript .gf-crawler-footer p { margin: 0; } @media (max-width: 700px) { noscript main { padding: 24px 16px; } noscript .gf-crawler-header-inner, noscript .gf-crawler-footer { padding-left: 16px; padding-right: 16px; } noscript .gf-crawler-topic-list-head { display: none; } noscript .gf-crawler-topic-row { display: grid; grid-template-columns: 1fr; padding: 18px 0; } noscript .gf-crawler-topic-row > div { width: 100%; padding: 0; text-align: left; white-space: normal; } noscript .gf-crawler-topic-row > div:not(:first-child) { width: auto; margin-right: 12px; color: #8b9098; font-size: 14px; white-space: nowrap; } noscript .gf-crawler-topic-title { font-size: 18px; } noscript .gf-crawler-topic-excerpt { margin-top: 8px; } noscript .gf-crawler-user-head { align-items: flex-start; } noscript .gf-crawler-user-stats dl { grid-template-columns: repeat(2, minmax(0, 1fr)); } }

YourTJ Forum

HACH合集 · 10月30日：AI 早报 + 日报

【HACH早报（2026-10-30）】 Mistral Large 4 权重正式开源：1 万亿参数，编程与防御安全领跑开源 Mistral 10 月 29 日放出改进版 Large 4（Le Chonk）权重：约 1 万亿总参数、520 亿激活参数、100 万 token 上下文，Hugging Face 可直接下载、无需审批，但要跑满血版需要数据中心级 GPU。授权为修改版 MIT，可免费商用...

技术分享 Gunpowder 2026-10-30T14:38:26Z 更新于 2026-10-30T14:38:42Z 0 回复 0 赞 8 浏览

帖子

#1 Gunpowder 2026-10-30T14:38:26Z
【HACH早报（2026-10-30）】
1. Mistral Large 4 权重正式开源：1 万亿参数，编程与防御安全领跑开源 Mistral 10 月 29 日放出改进版 Large 4（Le Chonk）权重：约 1 万亿总参数、520 亿激活参数、100 万 token 上下文，Hugging Face 可直接下载、无需审批，但要跑满血版需要数据中心级 GPU。授权为修改版 MIT，可免费商用、微调，只有对外以模型即服务运营才需另拿商业授权；官方 API 现价每百万 token 输入 0.70 美元、缓存输入 0.07 美元、输出 2.10 美元，首周还有 demo 免费额度可先体验。来源：https://pasqualepillitteri.it/en/news/28041/mistral-large-4-le-chonk-open-weights-hugging-face
2. 智元灵渠 OS 1.0 与 GO-3 Preview：具身方向的开源底座来了 10 月 29 日智元创新日上，首个具身智能原生操作系统灵渠 OS 1.0 发布并捐赠给开放原子开源基金会，支持 15 类 100 多种外设快速接入，1000Hz 高频任务下端到端时延抖动小于 4 微秒，计划 2026 年底全面开源；同场的新一代具身基座模型 GO-3 Preview（自研 WRAM 架构）登顶 RoboDojo、RoboTwin、RoboColiseum 三项评测，也计划年底全面开源。做具身的可以先盯住这个社区，年底权重一到就能上手。来源：https://feng.ifeng.com/c/8xcKho4Y6xk
3. Apex-35B-V1.0 开源：清华 27 岁助理教授团队的自进化研究模型超衍智能发布首个 35B 模型 Apex-35B-V1.0，Apache 2.0 协议，约 72GB BF16 权重另有 GGUF 量化版，Hugging Face 和 GitHub 都有。亮点是它由自动化研究系统 Apex Research 参与研发，兼顾通用任务（Coding、Agent、问答、深度搜索）和研究任务（AI 研究、计算科学、机器人仿真），想研究“AI 如何参与科研”可以直接拿来跑。来源：https://huggingface.co/Apexintelligence-AI/Apex-35B-V1.0
4. OpenCode 限时免费模型：免费，但只能在应用内用 OpenCode 的限时免费档包括 GLM-5.3-Flash、DeepSeek V4.1 Flash 等模型，免费但只能在 OpenCode agent 内使用，不能当通用 API 调。适合先在 agent 里白嫖体验，觉得好用再考虑付费或换自己的 Key。来源：https://hashnode.com/blog/free-open-weight-model-apis
5. Kilo Code：开源 coding agent，自带免 Key 的免费模式开源终端 coding agent Kilo Code 有 Auto Free 模式，走免 Key 的 OpenAI 兼容网关，每 IP 每小时 200 请求，免费模型池轮换（Nemotron、Laguna 等系列）。不用绑卡、不用填 Key 就能先跑起来，适合想零成本试 agent 编程的同学。来源：https://mvalentsev.github.io/awesome-free-ai-coding/
小提示：开源权重先看三点，许可证（能不能商用）、硬件门槛（要不要数据中心级显卡）、有没有免费 API 或 demo 可先体验再决定下载。

【HACH日报（2026-10-30）】
1. 智元发布具身基座模型GO-3（36氪）智元创新日推出GO-3、AGILE与WITA-OMNI，聚焦真实部署与有效智能工时度量。 https://finance.sina.cn/stock/jdts/2026-10-30/detail-iniwyncw5590337.d.html
2. OpenAI更新模型规范重塑对齐（环球网）新版规范以忠于人类意图为核心，要求模型理解真实诉求并接受人类监督。 https://finance.sina.com.cn/roll/2026-10-30/doc-iniwyncw5592708.shtml
3. 谷歌让大模型学会睡觉边学边记（新浪科技）睡眠框架以清醒与睡眠两阶段巩固记忆，缓解知识固化与灾难性遗忘问题。 https://finance.sina.com.cn/tech/digi/2026-10-30/doc-iniwysmu5539950.shtml
4. 韩国近半新模型基于中国基座（参考消息网）今年1至8月韩国新发布模型中49.3%基于中国模型，千问占比达46.8%。 https://finance.sina.cn/stock/jdts/2026-10-30/detail-iniwyncz8307193.d.html
5. 自动驾驶建筑公司Bedrock融资6亿美元（Analytics Insight） Bedrock完成6亿美元C轮融资，估值41亿美元，将加速自动驾驶工程机械商用部署。 https://www.analyticsinsight.net/news/startup-funding-round-up-ai-robotics-startups-raise-millions-this-week
HACH日报 X 主帖：https://x.com/GunpowderTwi/status/2116137354050355610

同济AI大模型交流群 Q 群号：1103021149，欢迎加入交流

回到首页 · 启用 JavaScript 以获得最佳体验。