HAT's Public Channel | 🫪

以及

不懂就问，今年是2021吗，不知道还以为用到 GPT-3.5-Turbo 了

火山 Coding Plan 真的是坑死了，文档对 AFP 与实际用量的换算逻辑描述得非常含糊不清，实际计算后 想用满额度只能保持 <32k ctx window。。。 将 <32k Context Window 作为 baseline, <64k 的额度变成六折 (~67%) >128k 额度变成三折 (33.5%) 并且不同模型、不同时间段的倍率也完全不一样，很多小巧思

以后可以用这模板了

Grok 真被 Cursor 完全夺舍成 CursorAI 了，最新模型甚至只能在 Cursor 和 Grok Build 里首发 SuperGrok 和 SuperGrok Heavy 成备胎🤡了

确认了，来自微信群小助手

DeepSeek V4 Pro 正式版已经更新至 API，调用模型名不变。新版本增强了 Agent 能力，支持 Responses API 和 Codex 接入，欢迎测试和反馈～

确认了，来自微信群小助手 纯网传，暂未找到来源

同样 0812 发的还有个 Grok 4.6 https://x.ai/news/grok-4-6

我 DS Harness 呢？我 DSv4 Flash 多模态呢？我 API 涨..哦这个算了

官网已更新！ DeepSeek-V4-Pro-0813 https://api-docs.deepseek.com/zh-cn/quick_start/pricing/

哦齁齁齁, DeepSeek V4 Pro 0812 好像推送到 API 了？ * _prod0820_fp8_kvcache_20260402 与 dsv4 flash 比较后应该是固定的 KV Cache Infra 后缀 Update: 确认更了 https://t.me/hatschannel/5782

傻孩子们，快跑啊—— Qwen3.8 2.4T 一改此前常用的 Apache 2.0 License, 而是选用了魔改的 "Qwen3.8-Max License" 开源协议，与 K3 全面看齐（甚至更甚）： 1. 当 产品/服务 > 1 亿 MAU，或月收入 > 2000 万美元时，UI 必须「显著展示对应的 Qwen 模型名称」 2. MaaS / AI “工作助手“ 的商业限制 · Model as a Service 服务商； · 或做独立的 AI Work Assistant (主要是 Coding Agent / 办公 AI)； · 且集团合计连续 12 个月收入 > 5000 万美元 ➡️ 那么，商业使用 Qwen3.8-Max 或其衍生模型前必须向 Qwen 取得额外许可。 原文:

Qwen3.8-Max License Copyright (c) 2026 Qwen Permission is hereby granted, free of charge, to any person obtaining a copy of this software, including the model weights, parameters, configuration files, inference code and associated documentation files (collectively, the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, sell, deploy, host, fine-tune, and create derivative works from (collectively, "Use" or "Using") copies of the Software; and to permit persons to whom the Software is furnished to do so, subject to the following conditions: 1. The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software. If the Software (or any derivative works thereof) is Used for any of the licensee's commercial products or services that have more than 100,000,000 monthly active users or US$ 20,000,000 (or equivalent in other currencies) monthly revenue, respective model name must be prominently displayed on the user interface of such product or service; and, 2. If the licensee or any of its affiliates conducts a Model as a Service or AI Work Assistant business, and the aggregate revenue of the licensee and its affiliates exceeds US$50,000,000 (or the equivalent amount in any other currencies) during any consecutive twelve (12) months, the licensee shall obtain a separate license from Qwen before Using the Software or its derivative works for any commercial purpose. The foregoing requirement shall not apply to the licensee's internal Use of the Software, provided that such Use does not make the Software, its outputs, or its underlying model capabilities available to any third party. "Model as a Service" means giving a third party access to language model inference or fine-tuning (e.g., via API or a hosted endpoint) in a manner that allows such third parties to exercise meaningful control over the inputs, parameters, or training data. This does not include the mere relaying of requests to models hosted by other third parties. “AI Work Assistant” means an independent AI-powered product primarily designed for AI-assisted coding or office productivity (e.g., Qoder and QwenWork). It does not include: (a) a single-purpose AI tool (such as an AI translation tool); (b) an AI assistant primarily designed for a domain other than coding or office productivity (such as Taobao AI Shopping Assistant or AMap AI Chat); or (c) an AI assistant that is a feature of a product whose primary purpose is not AI-assisted coding or office productivity. THE SOFTWARE AND ANY OUTPUT AND RESULTS THEREFROM ARE PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL QWEN, ITS AFFILIATES OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. THE USE OF THE SOFTWARE MUST COMPLY WITH APPLICABLE LAWS AND REGULATIONS, AND MUST NOT INFRINGE THE INTELLECTUAL PROPERTY RIGHTS OF ANY THIRD PARTY. For any questions regarding this license, please contact model-business@notice.qwencloud.com.

Qwen3.8 正式开源，但好像只先开了 2.4T-A95B, 27B 将在稍晚发布 Qwen3.8 支持 low/medium/xhigh 三档 effort level 可选，依旧默认 256k (272,166) 原生上下文、可扩展到 1M 由于频道主个人觉得 Qwen 系列的 Benchmark 没有什么实际意义，故本帖不附 Benchmark 图。 https://modelscope.cn/models/Qwen/Qwen3.8-2.4T-A95B https://huggingface.co/Qwen/Qwen3.8-2.4T-A95B X Post: https://x.com/Alibaba_Qwen/status/2084100707423289643

呃呃呃

微信官宣发布 WeLM-80B-A3B & WeLM-617B-A23B 其中, WeLM-80B-A3B 官宣代号 “Xiaowei” (小薇AI助手), 支持 "Native Weixin Functions" (吐槽: 连这都要自己创个新规范吗)、适用于 Chat & Search 场景。 WeLM-617B-A23B 则在此基础上支持了 “Xiaowei Tool Generation” 能力，并且仍在 “In Development”。 有意思的是，WeLM-80B-A3B 与 Qwen3-Next 总参数量和激活参数一模一样。 而 WeLM Blog 中则称这是他们用了不足 14T tokens 重训的，至于你们信不信，反正我是信了 Paper: https://github.com/Tencent/Sequential-Hidden-Decoding/~ Blog: WeLM-80B-A3B | WeLM-671B-A21B 配图来源

kimi code 飞书群弹消息还以为是有 Reset, 结果是这玩意？

K︀︁󠅄i︀︁󠄳m︀︁󠅇i︀︁󠅆 ︀︁︀更︀︁󠄇新︀︁󠅄订︀︁󠄳阅︀︁󠅇规︀︁󠅄则︀︁︁，︀︁󠄱订︀︁󠆎阅︀︁󠇇到︀︁󠄛期︀︁󠆌超︀︁󠅞过︀︁󠅚七︀︁󠅬天︀︁󠅘未︀︁󠆧续︀︁󠄋订︀︁󠇥的︀︁󠇩用︀︁󠇙户︀︁󠆰，︀︁󠅐将︀︁󠅁无︀︁󠄟法︀︁󠄼再︀︁󠅷次︀︁󠆌购︀︁󠅺买老套餐 订阅购买页更新公告: 续订规则调整：历史订阅用户可于 2026-XX-XX 00:00 前续订，逾期将无法直接购买

@所有人 各位 KimiCode 用户： K3 发布以来，感谢大家的热情支持。由于当前算力资源仍然紧张，为了优先保障正在订阅中的会员的使用体验，减少高峰期的排队与请求失败，我们将对老套餐的续订规则做如下调整： 自北京时间2026年8月20日起，老套餐「订阅过期 7 天以上」的用户无法续订。 即：订阅到期且超过 7 天未续订的账号，将无法再次购买老套餐。 ⚠️ 需要特别说明的是： - 正在订阅中的用户不受任何影响。 您的会员权益、自动续费均保持不变； - 自动续费正常开启的用户不受任何影响。 系统会照常扣费续期； - 本规则仅影响「订阅已到期且到期超过 7 天」的账号； - 公告时订阅会员已过期的历史用户，公告发布后7天内可买老套餐。 ❓FAQ： Q：我现在还在订阅中，受影响吗？ A：完全不受影响。在订用户权益不变，自动续费照常进行。 Q：我之前手动关了自动续费，到期后还能续老套餐吗？ A：可以，但请在到期后 7 天内完成续订。超过 7 天将无法再订阅老套餐。 Q：到期超过 7 天了，怎么办？ A：将无法再次购买老套餐。 Q：是不是每次订阅到期后，都有 7 天宽限期可以续费？忘了怎么办？ A：是的。每次到期后 7 天内均可续订老套餐；超过第 7 天仍未续订，将永久失去老套餐续订资格。 Q：老套餐还能升级档位吗？ A：可以，老套餐内升级（如 199 元档 → 699 元档）不受本次调整影响。 Q：这个 7 天的限制以后会变吗？ A：随着算力扩容，我们会持续评估并尽可能放宽限制，有变化会第一时间公告。

快讯: Mistral 现已在其官方平台中 self-hosted GLM-5.2, 并且将会支持越来越多的开源模型 欧洲原神。。。终于投降了吗 Read more: https://mistral.ai/news/regional-inference-open-models-new-compute/