<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
<channel>
<title>AI Index 资讯</title>
<link>https://ai.gotry.io/zh/news</link>
<atom:link href="https://ai.gotry.io/zh/news.xml" rel="self" type="application/rss+xml"/>
<description>来自厂商官方页面的 AI 模型与 API 资讯：发布、价格调整、下线和新能力。</description>
<language>zh-cn</language>
<lastBuildDate>Sat, 03 Oct 2026 00:00:00 GMT</lastBuildDate>
<item>
<title>xAI 停用 grok-voice-transcribe-1.0</title>
<link>https://ai.gotry.io/zh/news/2026-10-03-xai-retires-grok-voice-transcribe-1-0</link>
<guid isPermaLink="false">ai.gotry.io/news/2026-10-03-xai-retires-grok-voice-transcribe-1-0</guid>
<pubDate>Sat, 03 Oct 2026 00:00:00 GMT</pubDate>
<category>下线</category>
<category>xAI</category>
<description>grok-voice-transcribe-1.0 于 2026 年 10 月 2 日停止服务。发往该模型标识的请求会转至 grok-voice-transcribe-2.0，价格不变。 来源：xAI API release notes（https://docs.x.ai/developers/release-notes#grok-voice-transcribe-10-end-of-life）</description>
</item>
<item>
<title>Google 下线 Gemini 2.5 Flash Image</title>
<link>https://ai.gotry.io/zh/news/2026-10-02-google-retired-gemini-2-5-flash-image</link>
<guid isPermaLink="false">ai.gotry.io/news/2026-10-02-google-retired-gemini-2-5-flash-image</guid>
<pubDate>Fri, 02 Oct 2026 00:00:00 GMT</pubDate>
<category>下线</category>
<category>Google</category>
<description>Gemini 2.5 Flash Image 已下线。 来源：Release notes | Gemini API | Google AI for Developers（https://ai.google.dev/gemini-api/docs/changelog?hl=en）</description>
</item>
<item>
<title>OpenAI 弃用 gpt-4o-mini-tts (2025-12-15)、gpt-4o-mini-tts (2025-03-20)</title>
<link>https://ai.gotry.io/zh/news/2026-10-01-openai-deprecated-gpt-4o-mini-tts-2025-12-15-and-1-more</link>
<guid isPermaLink="false">ai.gotry.io/news/2026-10-01-openai-deprecated-gpt-4o-mini-tts-2025-12-15-and-1-more</guid>
<pubDate>Thu, 01 Oct 2026 00:00:00 GMT</pubDate>
<category>下线</category>
<category>OpenAI</category>
<description>gpt-4o-mini-tts (2025-12-15)、gpt-4o-mini-tts (2025-03-20) 已被弃用，将于 2027-01-06 下线。 来源：OpenAI API deprecations（https://developers.openai.com/api/docs/deprecations）</description>
</item>
<item>
<title>OpenAI 弃用 tts-1、tts-1-hd、GPT-5.4 nano、GPT-5.3 Codex、GPT-5.1</title>
<link>https://ai.gotry.io/zh/news/2026-10-01-openai-deprecated-tts-1-and-4-more</link>
<guid isPermaLink="false">ai.gotry.io/news/2026-10-01-openai-deprecated-tts-1-and-4-more</guid>
<pubDate>Thu, 01 Oct 2026 00:00:00 GMT</pubDate>
<category>下线</category>
<category>OpenAI</category>
<description>tts-1、tts-1-hd、GPT-5.4 nano、GPT-5.3 Codex、GPT-5.1 已被弃用，将于 2027-01-06 下线。 来源：OpenAI API deprecations（https://developers.openai.com/api/docs/deprecations）</description>
</item>
<item>
<title>OpenAI 下线 gpt-5.4-cyber</title>
<link>https://ai.gotry.io/zh/news/2026-10-01-openai-retired-gpt-5-4-cyber</link>
<guid isPermaLink="false">ai.gotry.io/news/2026-10-01-openai-retired-gpt-5-4-cyber</guid>
<pubDate>Thu, 01 Oct 2026 00:00:00 GMT</pubDate>
<category>下线</category>
<category>OpenAI</category>
<description>gpt-5.4-cyber 已下线。 来源：OpenAI API deprecations（https://developers.openai.com/api/docs/deprecations）</description>
</item>
<item>
<title>Alibaba Qwen 下线 deepseek-r1-distill-llama-8b</title>
<link>https://ai.gotry.io/zh/news/2026-09-30-alibaba-qwen-retired-deepseek-r1-distill-llama-8b</link>
<guid isPermaLink="false">ai.gotry.io/news/2026-09-30-alibaba-qwen-retired-deepseek-r1-distill-llama-8b</guid>
<pubDate>Wed, 30 Sep 2026 00:00:00 GMT</pubDate>
<category>下线</category>
<category>Alibaba Qwen</category>
<description>deepseek-r1-distill-llama-8b 已下线。 来源：Alibaba Cloud Model Studio: text generation models（https://www.alibabacloud.com/help/en/model-studio/text-generation-model）</description>
</item>
<item>
<title>Anthropic 计划停止 Claude Sonnet 4.5</title>
<link>https://ai.gotry.io/zh/news/2026-09-30-anthropic-schedules-claude-sonnet-4-5-retirement</link>
<guid isPermaLink="false">ai.gotry.io/news/2026-09-30-anthropic-schedules-claude-sonnet-4-5-retirement</guid>
<pubDate>Wed, 30 Sep 2026 00:00:00 GMT</pubDate>
<category>下线</category>
<category>Anthropic</category>
<description>Anthropic 计划于 2026 年 11 月 30 日在 Claude API 中停止提供 Claude Sonnet 4.5（claude-sonnet-4-5-20250929）。Anthropic 建议迁移至 Claude Sonnet 5.5。 来源：Claude Developer Platform release notes（https://platform.claude.com/docs/en/release-notes/overview#september-30-2026）</description>
</item>
<item>
<title>Black Forest Labs 发布 FLUX 3 Image</title>
<link>https://ai.gotry.io/zh/news/2026-09-30-black-forest-labs-releases-flux-3-image</link>
<guid isPermaLink="false">ai.gotry.io/news/2026-09-30-black-forest-labs-releases-flux-3-image</guid>
<pubDate>Wed, 30 Sep 2026 00:00:00 GMT</pubDate>
<category>发布</category>
<category>Black Forest Labs</category>
<description>FLUX 3 Image 已通过一个端点开放，可用于图像生成和编辑，最多支持十张参考图像，输出最高可达 4K。每张图像的价格从 768sq 分辨率的 0.041 美元到 4K 的 0.607 美元不等。 来源：BFL API release notes（https://docs.bfl.ai/release-notes#october-1-2026）</description>
</item>
<item>
<title>Google 发布 Gemini 4 Argon</title>
<link>https://ai.gotry.io/zh/news/2026-09-30-google-announces-gemini-4-argon</link>
<guid isPermaLink="false">ai.gotry.io/news/2026-09-30-google-announces-gemini-4-argon</guid>
<pubDate>Wed, 30 Sep 2026 00:00:00 GMT</pubDate>
<category>发布</category>
<category>Google</category>
<description>Google 发布 Gemini 4 Argon，输出 token 上限为 100 万。目前该模型正向可信的网络防御人员开放，开发者暂时无法使用；Google 表示之后将向开发者开放。 来源：Google DeepMind blog（https://deepmind.google/blog/gemini-4-argon-our-next-era-of-frontier-intelligence/）</description>
</item>
<item>
<title>Google 下线 gemini-omni-flash-preview</title>
<link>https://ai.gotry.io/zh/news/2026-09-30-google-retired-gemini-omni-flash-preview</link>
<guid isPermaLink="false">ai.gotry.io/news/2026-09-30-google-retired-gemini-omni-flash-preview</guid>
<pubDate>Wed, 30 Sep 2026 00:00:00 GMT</pubDate>
<category>下线</category>
<category>Google</category>
<description>gemini-omni-flash-preview 已下线。 来源：Gemini API deprecations（https://ai.google.dev/gemini-api/docs/deprecations?hl=en）</description>
</item>
<item>
<title>Runway 将 Eleven v4 引入 Runway Dev</title>
<link>https://ai.gotry.io/zh/news/2026-09-30-runway-adds-eleven-v4-to-runway-dev</link>
<guid isPermaLink="false">ai.gotry.io/news/2026-09-30-runway-adds-eleven-v4-to-runway-dev</guid>
<pubDate>Wed, 30 Sep 2026 00:00:00 GMT</pubDate>
<category>发布</category>
<category>Runway</category>
<description>Eleven v4 已在 Runway Dev 上提供语音生成，脚本最长可达 2,500 个字符。每 1,000 个字符在 2026 年 10 月 12 日 PT 前收费 2.2 credits，此后收费 5 credits，最低收费 1 credit。 来源：Runway API changelog（https://docs.dev.runwayml.com/api-details/api_changelog/#eleven-v4-on-runway-dev）</description>
</item>
<item>
<title>Inception 正式推出 Mercury Voice</title>
<link>https://ai.gotry.io/zh/news/2026-09-29-inception-makes-mercury-voice-generally-available</link>
<guid isPermaLink="false">ai.gotry.io/news/2026-09-29-inception-makes-mercury-voice-generally-available</guid>
<pubDate>Tue, 29 Sep 2026 00:00:00 GMT</pubDate>
<category>发布</category>
<category>Inception</category>
<description>Inception 已面向企业客户正式推出 Mercury Voice，支持 128K token 上下文、最多 50K 输出 token 和三档推理设置。标准价为每百万输入 token 0.40 美元、每百万输出 token 1.50 美元，发布期间五折（0.20 美元与 0.75 美元）。 来源：Inception blog（https://www.inceptionlabs.ai/blog/introducing-mercury-voice）</description>
</item>
<item>
<title>MiniMax 发布 minimax-m3.1-flash-preview</title>
<link>https://ai.gotry.io/zh/news/2026-09-29-minimax-m3-1-flash-preview-release</link>
<guid isPermaLink="false">ai.gotry.io/news/2026-09-29-minimax-m3-1-flash-preview-release</guid>
<pubDate>Tue, 29 Sep 2026 00:00:00 GMT</pubDate>
<category>发布</category>
<category>MiniMax</category>
<description>上下文 1M tokens。 来源：Subscription Plan Upgrade Announcement - MiniMax API Docs（https://platform.minimax.io/docs/token-plan/announcements）</description>
</item>
<item>
<title>Mistral AI 停用 Leanstral 1.5 和 GLM 5.2</title>
<link>https://ai.gotry.io/zh/news/2026-09-29-mistral-retires-leanstral-1-5-and-glm-5-2</link>
<guid isPermaLink="false">ai.gotry.io/news/2026-09-29-mistral-retires-leanstral-1-5-and-glm-5-2</guid>
<pubDate>Tue, 29 Sep 2026 00:00:00 GMT</pubDate>
<category>下线</category>
<category>Mistral AI</category>
<description>Leanstral 1.5 将于 2026 年 9 月 30 日停用；Z.ai GLM 5.2 将于 2026 年 10 月 31 日停用。Mistral AI 建议以同价的 Z.ai GLM 5.3 替代后者。 来源：Mistral AI changelog（https://docs.mistral.ai/resources/changelogs#date-2026-09-29）</description>
</item>
<item>
<title>OpenAI 为 Agents API 添加计算机使用能力</title>
<link>https://ai.gotry.io/zh/news/2026-09-29-openai-adds-computer-use-to-the-agents-api</link>
<guid isPermaLink="false">ai.gotry.io/news/2026-09-29-openai-adds-computer-use-to-the-agents-api</guid>
<pubDate>Tue, 29 Sep 2026 00:00:00 GMT</pubDate>
<category>能力</category>
<category>OpenAI</category>
<description>OpenAI 为 Agents API 添加了计算机使用能力，允许智能体在 OpenAI 托管的浏览器中完成任务。网站访问审批和登录由应用程序处理。 来源：OpenAI API changelog（https://developers.openai.com/api/docs/changelog）</description>
</item>
<item>
<title>OpenAI 为 GPT-6 Astra 添加 Ultrafast 模式</title>
<link>https://ai.gotry.io/zh/news/2026-09-29-openai-adds-ultrafast-mode-for-gpt-6-astra</link>
<guid isPermaLink="false">ai.gotry.io/news/2026-09-29-openai-adds-ultrafast-mode-for-gpt-6-astra</guid>
<pubDate>Tue, 29 Sep 2026 00:00:00 GMT</pubDate>
<category>能力</category>
<category>OpenAI</category>
<description>OpenAI 在 Responses API 中为 GPT-6 Astra 添加了 Ultrafast 模式。API 客户可使用 ultrafast 服务层缩短生成输出令牌之间的时间，但受速率限制约束。 来源：OpenAI API changelog（https://developers.openai.com/api/docs/changelog）</description>
</item>
<item>
<title>OpenAI 发布 GPT-6.1 Sol</title>
<link>https://ai.gotry.io/zh/news/2026-09-29-openai-releases-gpt-6-1-sol</link>
<guid isPermaLink="false">ai.gotry.io/news/2026-09-29-openai-releases-gpt-6-1-sol</guid>
<pubDate>Tue, 29 Sep 2026 00:00:00 GMT</pubDate>
<category>发布</category>
<category>OpenAI</category>
<description>OpenAI 发布了 GPT-6.1 Sol（gpt-6.1-sol），用于复杂编程和专业工作。输入不超过 272K token 的提示按标准价计费：每百万输入 token 2 美元、缓存输入 0.10 美元、输出 10 美元。 来源：OpenAI API changelog（https://developers.openai.com/api/docs/changelog）</description>
</item>
<item>
<title>Anthropic 发布 Claude Sonnet 5.5</title>
<link>https://ai.gotry.io/zh/news/2026-09-28-anthropic-releases-claude-sonnet-5-5</link>
<guid isPermaLink="false">ai.gotry.io/news/2026-09-28-anthropic-releases-claude-sonnet-5-5</guid>
<pubDate>Mon, 28 Sep 2026 00:00:00 GMT</pubDate>
<category>发布</category>
<category>Anthropic</category>
<description>Anthropic 推出 Claude Sonnet 5.5（claude-sonnet-5-5），已在 Claude API、Amazon Bedrock、Claude Platform on AWS、Google Cloud 和 Microsoft Foundry 提供。上下文窗口、输出上限与价格见其模型页面。 来源：Claude Developer Platform release notes（https://platform.claude.com/docs/en/release-notes/overview#september-28-2026）</description>
</item>
<item>
<title>ElevenLabs 发布 Eleven v4 和 Eleven v4 Turbo</title>
<link>https://ai.gotry.io/zh/news/2026-09-28-elevenlabs-releases-eleven-v4-and-eleven-v4-turbo</link>
<guid isPermaLink="false">ai.gotry.io/news/2026-09-28-elevenlabs-releases-eleven-v4-and-eleven-v4-turbo</guid>
<pubDate>Mon, 28 Sep 2026 00:00:00 GMT</pubDate>
<category>发布</category>
<category>ElevenLabs</category>
<description>Eleven v4 和 Eleven v4 Turbo 现已推出。Eleven v4 支持 90 多种语言的语音克隆；Turbo 面向实时应用，推理延迟中位数约为 100 毫秒。 来源：ElevenLabs changelog（https://elevenlabs.io/docs/changelog/2026/9/28）</description>
</item>
<item>
<title>Meituan 发布 LongCat-2.5-Preview</title>
<link>https://ai.gotry.io/zh/news/2026-09-25-longcat-2-5-preview-release</link>
<guid isPermaLink="false">ai.gotry.io/news/2026-09-25-longcat-2-5-preview-release</guid>
<pubDate>Fri, 25 Sep 2026 00:00:00 GMT</pubDate>
<category>发布</category>
<category>Meituan</category>
<description>上下文 1M tokens。LongCat API：输入 $0.30、输出 $1.20 每百万 tokens。 来源：LongCat API: changelog（https://longcat.chat/platform/docs/ChangeLog.html）</description>
</item>
<item>
<title>Perplexity 下调 fast preset 搜索价格</title>
<link>https://ai.gotry.io/zh/news/2026-09-25-perplexity-cuts-fast-preset-search-price</link>
<guid isPermaLink="false">ai.gotry.io/news/2026-09-25-perplexity-cuts-fast-preset-search-price</guid>
<pubDate>Fri, 25 Sep 2026 00:00:00 GMT</pubDate>
<category>价格</category>
<category>Perplexity</category>
<description>Agent API 的 fast preset 现改用 Fast Search。web_search 从每 1,000 次调用 $2.50 降至 $1.00，搜索约快 800 毫秒。 来源：Perplexity API changelog（https://docs.perplexity.ai/docs/resources/changelog#september-2026）</description>
</item>
<item>
<title>Perplexity 新增 Fast Search 选项</title>
<link>https://ai.gotry.io/zh/news/2026-09-24-perplexity-adds-fast-search-option</link>
<guid isPermaLink="false">ai.gotry.io/news/2026-09-24-perplexity-adds-fast-search-option</guid>
<pubDate>Thu, 24 Sep 2026 00:00:00 GMT</pubDate>
<category>能力</category>
<category>Perplexity</category>
<description>Search API 与 Agent API 的 web_search 可将 search_type 设为 fast，走更低延迟路径，每 1,000 次请求或调用 $1.00。Agent API 的模型 token 另计；search_type web 为标准网页搜索。 来源：Perplexity API changelog（https://docs.perplexity.ai/docs/resources/changelog#september-2026）</description>
</item>
<item>
<title>Black Forest Labs 发布 FLUX 3 Action</title>
<link>https://ai.gotry.io/zh/news/2026-09-23-black-forest-labs-releases-flux-3-action</link>
<guid isPermaLink="false">ai.gotry.io/news/2026-09-23-black-forest-labs-releases-flux-3-action</guid>
<pubDate>Wed, 23 Sep 2026 00:00:00 GMT</pubDate>
<category>发布</category>
<category>Black Forest Labs</category>
<description>Black Forest Labs 发布开源权重的 7B 世界-动作模型 FLUX 3 Action，基于其 FLUX 3 骨干，并提供权重下载。在 RoboLab-120 上，参数不到此前最佳开源模型的一半，速度最高达 3.95 倍。 来源：Black Forest Labs blog（https://bfl.ai/blog/flux-3-action）</description>
</item>
<item>
<title>Meta 推出 Muse Realtime Avatar</title>
<link>https://ai.gotry.io/zh/news/2026-09-23-meta-introduces-muse-realtime-avatar</link>
<guid isPermaLink="false">ai.gotry.io/news/2026-09-23-meta-introduces-muse-realtime-avatar</guid>
<pubDate>Wed, 23 Sep 2026 00:00:00 GMT</pubDate>
<category>发布</category>
<category>Meta</category>
<description>2026 年 9 月 23 日，Meta 推出 Muse Realtime Avatar。它依据 Muse Realtime Voice 的语音 token 与参考素材，实时生成同步的面部、手部与全身视频。 来源：Meta Superintelligence Labs research（https://research.meta.ai/blog/bringing-your-muse-to-life）</description>
</item>
<item>
<title>Anthropic 支持在对话中途定义工具</title>
<link>https://ai.gotry.io/zh/news/2026-09-22-anthropic-adds-mid-conversation-tool-definitions</link>
<guid isPermaLink="false">ai.gotry.io/news/2026-09-22-anthropic-adds-mid-conversation-tool-definitions</guid>
<pubDate>Tue, 22 Sep 2026 00:00:00 GMT</pubDate>
<category>能力</category>
<category>Anthropic</category>
<description>在 Claude API 上，借助 inline-tools-2026-09-15 beta 头，可在对话中途的系统消息中放入完整工具定义；配合 MCP connector beta 头还可使用 MCP 工具集，而无需修改 tools 列表。 来源：Claude Developer Platform release notes（https://platform.claude.com/docs/en/release-notes/overview#september-22-2026）</description>
</item>
<item>
<title>Anthropic 为 Claude Opus 5.5 推出快速模式预览</title>
<link>https://ai.gotry.io/zh/news/2026-09-22-anthropic-previews-fast-mode-for-claude-opus-5-5</link>
<guid isPermaLink="false">ai.gotry.io/news/2026-09-22-anthropic-previews-fast-mode-for-claude-opus-5-5</guid>
<pubDate>Tue, 22 Sep 2026 00:00:00 GMT</pubDate>
<category>能力</category>
<category>Anthropic</category>
<description>Claude API 上的 Claude Opus 5.5 现提供研究预览版快速模式。 来源：Claude Developer Platform release notes（https://platform.claude.com/docs/en/release-notes/overview#september-22-2026）</description>
</item>
<item>
<title>Anthropic 发布 Claude Opus 5.5</title>
<link>https://ai.gotry.io/zh/news/2026-09-22-anthropic-releases-claude-opus-5-5</link>
<guid isPermaLink="false">ai.gotry.io/news/2026-09-22-anthropic-releases-claude-opus-5-5</guid>
<pubDate>Tue, 22 Sep 2026 00:00:00 GMT</pubDate>
<category>发布</category>
<category>Anthropic</category>
<description>Anthropic 发布 Claude Opus 5.5（claude-opus-5-5）：100 万 token 上下文、最大输出 12.8 万 token，自适应思考始终开启，价格为每百万 token $4/$20（Claude Opus 5 为 $5/$25）。已在 Claude API、Amazon Bedrock、AWS、Google Cloud 和 Microsoft Foundry 提供。 来源：Claude Developer Platform release notes（https://platform.claude.com/docs/en/release-notes/overview#september-22-2026）</description>
</item>
<item>
<title>Google 发布 Gemini 3.8 Flash TTS 模型</title>
<link>https://ai.gotry.io/zh/news/2026-09-22-google-releases-gemini-3-8-flash-tts-models</link>
<guid isPermaLink="false">ai.gotry.io/news/2026-09-22-google-releases-gemini-3-8-flash-tts-models</guid>
<pubDate>Tue, 22 Sep 2026 00:00:00 GMT</pubDate>
<category>发布</category>
<category>Google</category>
<description>Gemini 3.8 Flash TTS（gemini-3.8-flash-tts）与 Gemini 3.8 Flash-Lite TTS（gemini-3.8-flash-lite-tts）正式可用，并提供 Gemini API Voices 端点。Flash-Lite TTS 旨在取代 gemini-3.1-flash-tts-preview。 来源：Gemini API changelog（https://ai.google.dev/gemini-api/docs/changelog#09-22-2026）</description>
</item>
<item>
<title>Xiaomi 发布 MiMo-V2.6-Flash</title>
<link>https://ai.gotry.io/zh/news/2026-09-22-mimo-v2-6-flash-release</link>
<guid isPermaLink="false">ai.gotry.io/news/2026-09-22-mimo-v2-6-flash-release</guid>
<pubDate>Tue, 22 Sep 2026 00:00:00 GMT</pubDate>
<category>发布</category>
<category>Xiaomi</category>
<description>上下文 1M tokens。Xiaomi MiMo API：输入 $0.14、输出 $0.28 每百万 tokens。 来源：Xiaomi MiMo: model updates（https://mimo.mi.com/docs/en-US/updates/model）</description>
</item>
<item>
<title>Xiaomi 发布 MiMo-V2.6-Pro</title>
<link>https://ai.gotry.io/zh/news/2026-09-22-mimo-v2-6-pro-release</link>
<guid isPermaLink="false">ai.gotry.io/news/2026-09-22-mimo-v2-6-pro-release</guid>
<pubDate>Tue, 22 Sep 2026 00:00:00 GMT</pubDate>
<category>发布</category>
<category>Xiaomi</category>
<description>上下文 1M tokens。Xiaomi MiMo API：输入 $0.435、输出 $0.87 每百万 tokens。 来源：Xiaomi MiMo: model updates（https://mimo.mi.com/docs/en-US/updates/model）</description>
</item>
<item>
<title>Xiaomi 发布 MiMo-V2.6-Pro-UltraSpeed</title>
<link>https://ai.gotry.io/zh/news/2026-09-22-mimo-v2-6-pro-ultraspeed-release</link>
<guid isPermaLink="false">ai.gotry.io/news/2026-09-22-mimo-v2-6-pro-ultraspeed-release</guid>
<pubDate>Tue, 22 Sep 2026 00:00:00 GMT</pubDate>
<category>发布</category>
<category>Xiaomi</category>
<description>上下文 1M tokens。Xiaomi MiMo API：输入 $4.35、输出 $8.70 每百万 tokens。 来源：Xiaomi MiMo: model updates（https://mimo.mi.com/docs/en-US/updates/model）</description>
</item>
<item>
<title>OpenAI 改进 GPT-6 的提示缓存</title>
<link>https://ai.gotry.io/zh/news/2026-09-22-openai-improves-prompt-caching-for-gpt-6</link>
<guid isPermaLink="false">ai.gotry.io/news/2026-09-22-openai-improves-prompt-caching-for-gpt-6</guid>
<pubDate>Tue, 22 Sep 2026 00:00:00 GMT</pubDate>
<category>能力</category>
<category>OpenAI</category>
<description>OpenAI 表示，GPT-6 系列默认的提示缓存命中率更高，30 分钟内重复使用的共享前缀可享缓存折扣。新增工具包括缓存用量面板、解释缓存未命中原因的诊断工具、显式缓存断点，以及在对话中途调整推理强度而不破坏缓存。 来源：OpenAI news（https://openai.com/index/better-prompt-caching-for-gpt-6）</description>
</item>
<item>
<title>OpenAI 发布 GPT-6 Sol 与 GPT-6 Luna</title>
<link>https://ai.gotry.io/zh/news/2026-09-22-openai-releases-gpt-6-sol-and-gpt-6-luna</link>
<guid isPermaLink="false">ai.gotry.io/news/2026-09-22-openai-releases-gpt-6-sol-and-gpt-6-luna</guid>
<pubDate>Tue, 22 Sep 2026 00:00:00 GMT</pubDate>
<category>发布</category>
<category>OpenAI</category>
<description>OpenAI 发布 GPT-6 Sol（gpt-6-sol）与 GPT-6 Luna（gpt-6-luna），可通过 Responses 与 Chat Completions 接收文本和图像并生成文本。输入不超过 272K token 的提示按标准价计费，每百万 token：Sol 输入 $2、缓存输入 $0.20、输出 $10；Luna 为 $0.10、$0.01 与 $0.50。 来源：OpenAI API changelog（https://developers.openai.com/api/docs/changelog）</description>
</item>
<item>
<title>xAI 发布 Grok 4.7</title>
<link>https://ai.gotry.io/zh/news/2026-09-21-grok-4-7-release</link>
<guid isPermaLink="false">ai.gotry.io/news/2026-09-21-grok-4-7-release</guid>
<pubDate>Mon, 21 Sep 2026 00:00:00 GMT</pubDate>
<category>发布</category>
<category>xAI</category>
<description>上下文 500K tokens。xAI API：输入 $2.00、输出 $6.00 每百万 tokens。 来源：xAI release notes（https://docs.x.ai/developers/release-notes）</description>
</item>
<item>
<title>Alibaba Qwen 发布 qwen3.8-omni-flash-realtime</title>
<link>https://ai.gotry.io/zh/news/2026-09-21-qwen3-8-omni-flash-realtime-release</link>
<guid isPermaLink="false">ai.gotry.io/news/2026-09-21-qwen3-8-omni-flash-realtime-release</guid>
<pubDate>Mon, 21 Sep 2026 00:00:00 GMT</pubDate>
<category>发布</category>
<category>Alibaba Qwen</category>
<description>上下文 197K tokens。 来源：Alibaba Cloud Model Studio: newly released models（https://www.alibabacloud.com/help/en/model-studio/newly-released-models）</description>
</item>
<item>
<title>Qwen 开源 Qwen-Image-2.1</title>
<link>https://ai.gotry.io/zh/news/2026-09-20-alibaba-releases-qwen-image-2-1</link>
<guid isPermaLink="false">ai.gotry.io/news/2026-09-20-alibaba-releases-qwen-image-2-1</guid>
<pubDate>Sun, 20 Sep 2026 00:00:00 GMT</pubDate>
<category>发布</category>
<category>Alibaba Qwen</category>
<description>Qwen 于 2026 年 9 月 20 日开源 Qwen-Image-2.1。这款拥有 70 亿参数的图像模型融合了文生图和图像编辑，并原生支持透明图像。 来源：Qwen research（https://qwen.ai/blog?id=qwen-image-2.1）</description>
</item>
<item>
<title>Alibaba Qwen 发布 qwen-audio-3.1-realtime-plus</title>
<link>https://ai.gotry.io/zh/news/2026-09-20-qwen-audio-3-1-realtime-plus-release</link>
<guid isPermaLink="false">ai.gotry.io/news/2026-09-20-qwen-audio-3-1-realtime-plus-release</guid>
<pubDate>Sun, 20 Sep 2026 00:00:00 GMT</pubDate>
<category>发布</category>
<category>Alibaba Qwen</category>
<description>上下文 262K tokens。 来源：Alibaba Cloud Model Studio: newly released models（https://www.alibabacloud.com/help/en/model-studio/newly-released-models）</description>
</item>
<item>
<title>Qwen 发布 Qwen3.8-LiveTranslate</title>
<link>https://ai.gotry.io/zh/news/2026-09-18-alibaba-releases-qwen3-8-livetranslate</link>
<guid isPermaLink="false">ai.gotry.io/news/2026-09-18-alibaba-releases-qwen3-8-livetranslate</guid>
<pubDate>Fri, 18 Sep 2026 00:00:00 GMT</pubDate>
<category>发布</category>
<category>Alibaba Qwen</category>
<description>Qwen 于 2026 年 9 月 18 日发布 Qwen3.8-LiveTranslate，称其采用 Interleave 架构进行实时同声传译。其平均滞后指标从 2.8 秒降至 2.3 秒。 来源：Qwen research（https://qwen.ai/blog?id=qwen3.8-livetranslate）</description>
</item>
<item>
<title>Google 限制 Gemini 2.5 模型访问</title>
<link>https://ai.gotry.io/zh/news/2026-09-18-google-limits-access-to-gemini-2-5-models</link>
<guid isPermaLink="false">ai.gotry.io/news/2026-09-18-google-limits-access-to-gemini-2-5-models</guid>
<pubDate>Fri, 18 Sep 2026 00:00:00 GMT</pubDate>
<category>其他</category>
<category>Google</category>
<description>Google 将 Gemini 2.5 的 API 访问限制为曾实际使用过这些模型的用户。模型未弃用，在另行通知前仍可通过 API 提供。新项目应使用 3.5 Flash-Lite 或 3.8 Flash。 来源：Gemini API changelog（https://ai.google.dev/gemini-api/docs/changelog#09-18-2026）</description>
</item>
<item>
<title>Google 发布 Antigravity Agent 09-2026</title>
<link>https://ai.gotry.io/zh/news/2026-09-17-google-releases-antigravity-agent-09-2026</link>
<guid isPermaLink="false">ai.gotry.io/news/2026-09-17-google-releases-antigravity-agent-09-2026</guid>
<pubDate>Thu, 17 Sep 2026 00:00:00 GMT</pubDate>
<category>发布</category>
<category>Google</category>
<description>Google 发布 antigravity-preview-09-2026，取代 antigravity-preview-05-2026。本地工具参数与文件编辑有变；仅读取远程输出时只需更新 agent 字符串。5 月预览版将于 2026 年 10 月 5 日关闭。 来源：Gemini API changelog（https://ai.google.dev/gemini-api/docs/changelog#09-17-2026）</description>
</item>
<item>
<title>Alibaba Qwen 发布 qwen3.8-omni-flash</title>
<link>https://ai.gotry.io/zh/news/2026-09-17-qwen3-8-omni-flash-release</link>
<guid isPermaLink="false">ai.gotry.io/news/2026-09-17-qwen3-8-omni-flash-release</guid>
<pubDate>Thu, 17 Sep 2026 00:00:00 GMT</pubDate>
<category>发布</category>
<category>Alibaba Qwen</category>
<description>上下文 1M tokens。Alibaba Cloud Model Studio：输入 $0.15、输出 $0.47 每百万 tokens。 来源：模型上下架与更新-大模型服务平台百炼-阿里云_大模型服务平台百炼(Model Studio)-大模型服务平台百炼(Model Studio)-阿里云帮助中心（https://help.aliyun.com/zh/model-studio/newly-released-models）</description>
</item>
<item>
<title>Runway 在 Dev 上线 Enhance Frame Rate</title>
<link>https://ai.gotry.io/zh/news/2026-09-17-runway-adds-enhance-frame-rate-on-dev</link>
<guid isPermaLink="false">ai.gotry.io/news/2026-09-17-runway-adds-enhance-frame-rate-on-dev</guid>
<pubDate>Thu, 17 Sep 2026 00:00:00 GMT</pubDate>
<category>能力</category>
<category>Runway</category>
<description>Runway Dev 现可将视频转换到 24、25、30、48、50、60、120、23.98、29.97 或 59.94 fps。输入最长 300 秒，按每 2 秒 1 credit 计费，通过视频放大端点并指定模型 enhance_frame_rate 使用。 来源：Runway API changelog（https://docs.dev.runwayml.com/api-details/api_changelog/#enhance-frame-rate-on-runway-dev）</description>
</item>
<item>
<title>Alibaba Qwen 发布 qwen-mt-uni</title>
<link>https://ai.gotry.io/zh/news/2026-09-16-qwen-mt-uni-release</link>
<guid isPermaLink="false">ai.gotry.io/news/2026-09-16-qwen-mt-uni-release</guid>
<pubDate>Wed, 16 Sep 2026 00:00:00 GMT</pubDate>
<category>发布</category>
<category>Alibaba Qwen</category>
<description>qwen-mt-uni 已上线。 来源：模型上下架与更新-大模型服务平台百炼-阿里云_大模型服务平台百炼(Model Studio)-大模型服务平台百炼(Model Studio)-阿里云帮助中心（https://help.aliyun.com/zh/model-studio/newly-released-models）</description>
</item>
<item>
<title>Google 发布 Gemini 3.8 Live 模型</title>
<link>https://ai.gotry.io/zh/news/2026-09-15-google-releases-gemini-3-8-live-models</link>
<guid isPermaLink="false">ai.gotry.io/news/2026-09-15-google-releases-gemini-3-8-live-models</guid>
<pubDate>Tue, 15 Sep 2026 00:00:00 GMT</pubDate>
<category>发布</category>
<category>Google</category>
<description>Gemini 3.8 Live（gemini-3.8-live）与 Gemini 3.8 Live Extended Thinking（gemini-3.8-live-extended-thinking）正式可用，为 Live API 上的音频到音频模型。 来源：Gemini API changelog（https://ai.google.dev/gemini-api/docs/changelog#09-15-2026）</description>
</item>
<item>
<title>TypeSafe 发布 Jev（预览版）</title>
<link>https://ai.gotry.io/zh/news/2026-09-15-jev-release</link>
<guid isPermaLink="false">ai.gotry.io/news/2026-09-15-jev-release</guid>
<pubDate>Tue, 15 Sep 2026 00:00:00 GMT</pubDate>
<category>发布</category>
<category>TypeSafe</category>
<description>上下文 64K tokens。TypeSafe API：输入 $0.042、输出 $0.00 每百万 tokens。 来源：Introducing System One models &amp; Jev（https://typesafe.ai/blog/introducing-system-one-models-and-jev）</description>
</item>
<item>
<title>Kling AI 发布 Virtual Try-On 3.0</title>
<link>https://ai.gotry.io/zh/news/2026-09-15-kuaishou-releases-virtual-try-on-3-0</link>
<guid isPermaLink="false">ai.gotry.io/news/2026-09-15-kuaishou-releases-virtual-try-on-3-0</guid>
<pubDate>Tue, 15 Sep 2026 00:00:00 GMT</pubDate>
<category>发布</category>
<category>Kuaishou Kling</category>
<description>Virtual Try-On 3.0 提升人脸一致性与画面质量。支持平铺、人台和上身服装图，并可锁定姿势、保留人脸以及选择是否保留原背景。V1 与 V1.5 参数已适配到 3.0 流程。 来源：Kling AI API updates（https://kling.ai/document-api/updates/api）</description>
</item>
<item>
<title>Anthropic 为 Messages API 增加按需对话压缩</title>
<link>https://ai.gotry.io/zh/news/2026-09-14-anthropic-adds-on-demand-compaction-to-messages-api</link>
<guid isPermaLink="false">ai.gotry.io/news/2026-09-14-anthropic-adds-on-demand-compaction-to-messages-api</guid>
<pubDate>Mon, 14 Sep 2026 00:00:00 GMT</pubDate>
<category>能力</category>
<category>Anthropic</category>
<description>Anthropic 的 Messages API 现以 beta（compact-2026-09-04）支持按需压缩对话。发送 compaction 参数后返回已签名摘要块，后续请求可将其置于前面以替代原消息，并保留最近几轮原文。 来源：Claude Developer Platform release notes（https://platform.claude.com/docs/en/release-notes/overview#september-14-2026）</description>
</item>
<item>
<title>Perplexity 支持自定义 MCP connector</title>
<link>https://ai.gotry.io/zh/news/2026-09-13-perplexity-adds-custom-mcp-connectors</link>
<guid isPermaLink="false">ai.gotry.io/news/2026-09-13-perplexity-adds-custom-mcp-connectors</guid>
<pubDate>Sun, 13 Sep 2026 00:00:00 GMT</pubDate>
<category>能力</category>
<category>Perplexity</category>
<description>Project 只需注册一次远程 MCP 服务器，凭证由 Perplexity 保存。Agent API 请求使用 connector ID，type 为 connector。支持 API key 或无认证，以及 Streamable HTTP 或 SSE。 来源：Perplexity API changelog（https://docs.perplexity.ai/docs/resources/changelog#september-2026）</description>
</item>
<item>
<title>ElevenLabs 正式推出 Scribe v2 Medical</title>
<link>https://ai.gotry.io/zh/news/2026-09-11-elevenlabs-makes-scribe-v2-medical-generally-available</link>
<guid isPermaLink="false">ai.gotry.io/news/2026-09-11-elevenlabs-makes-scribe-v2-medical-generally-available</guid>
<pubDate>Fri, 11 Sep 2026 00:00:00 GMT</pubDate>
<category>发布</category>
<category>ElevenLabs</category>
<description>Scribe v2 Medical 现已正式开放，用于批量医疗和临床语音识别。其收费与 Scribe v2 相同。 来源：ElevenLabs changelog（https://elevenlabs.io/docs/changelog/2026/9/11）</description>
</item>
<item>
<title>Black Forest Labs 为 FLUX 3 Video 增加 2K 与 4K</title>
<link>https://ai.gotry.io/zh/news/2026-09-10-black-forest-labs-adds-2k-and-4k-flux-3-video</link>
<guid isPermaLink="false">ai.gotry.io/news/2026-09-10-black-forest-labs-adds-2k-and-4k-flux-3-video</guid>
<pubDate>Thu, 10 Sep 2026 00:00:00 GMT</pubDate>
<category>能力</category>
<category>Black Forest Labs</category>
<description>FLUX 3 Video 接口现可一次返回 qhd（2560×1440）与 uhd（3840×2176），draft_enhance 也可将草稿直接提升到这两种分辨率（草稿本身仍为 hd）。文生与图生每秒加收 $0.40 与 $0.80，视频续写为 $0.65 与 $0.95。 来源：BFL API release notes（https://docs.bfl.ai/release-notes#september-10-2026）</description>
</item>
<item>
<title>DeepSeek 发布 DeepSeek-V4.1-Flash</title>
<link>https://ai.gotry.io/zh/news/2026-09-10-deepseek-releases-deepseek-v4-1-flash</link>
<guid isPermaLink="false">ai.gotry.io/news/2026-09-10-deepseek-releases-deepseek-v4-1-flash</guid>
<pubDate>Thu, 10 Sep 2026 00:00:00 GMT</pubDate>
<category>发布</category>
<category>DeepSeek</category>
<description>DeepSeek 发布新架构系列中最小的 DeepSeek-V4.1-Flash，具备原生多模态视觉理解，API 模型名为 deepseek-flash。已退役的 deepseek-v4-flash 与 deepseek-v4-flash-vision-exp 暂时路由到该模型。V4 Pro 在 2026 年 9 月 14 日后继续提供且计费不变，API 价格随本次发布下调。 来源：DeepSeek API change log（https://api-docs.deepseek.com/updates）</description>
</item>
<item>
<title>OpenAI 正式推出 GPT-Live 1</title>
<link>https://ai.gotry.io/zh/news/2026-09-10-openai-makes-gpt-live-1-generally-available</link>
<guid isPermaLink="false">ai.gotry.io/news/2026-09-10-openai-makes-gpt-live-1-generally-available</guid>
<pubDate>Thu, 10 Sep 2026 00:00:00 GMT</pubDate>
<category>发布</category>
<category>OpenAI</category>
<description>GPT-Live 1 已在 API 中正式可用，支持全双工语音会话，后端模型或智能体可同时处理推理与工具。语音按每分钟 $0.05、按秒计费；模型与工具用量另行收费。 来源：OpenAI API changelog（https://developers.openai.com/api/docs/changelog）</description>
</item>
<item>
<title>OpenAI 公开发布 Agents API 测试版</title>
<link>https://ai.gotry.io/zh/news/2026-09-10-openai-releases-agents-api-in-public-beta</link>
<guid isPermaLink="false">ai.gotry.io/news/2026-09-10-openai-releases-agents-api-in-public-beta</guid>
<pubDate>Thu, 10 Sep 2026 00:00:00 GMT</pubDate>
<category>能力</category>
<category>OpenAI</category>
<description>OpenAI 以公开测试版发布 Agents API，由托管的 Codex 运行环境负责会话编排、上下文压缩与恢复。会话可流式返回进度，接入自有工具与 MCP 服务器，并在 OpenAI 或客户沙箱中运行。 来源：OpenAI API changelog（https://developers.openai.com/api/docs/changelog）</description>
</item>
<item>
<title>Inception 发布 Mercury 2.5</title>
<link>https://ai.gotry.io/zh/news/2026-09-08-inception-releases-mercury-2-5</link>
<guid isPermaLink="false">ai.gotry.io/news/2026-09-08-inception-releases-mercury-2-5</guid>
<pubDate>Tue, 08 Sep 2026 00:00:00 GMT</pubDate>
<category>发布</category>
<category>Inception</category>
<description>Inception 发布 Mercury 2.5，质量高于 Mercury 2，上下文为 260K tokens，速度为每秒 1,107 tokens。标价为每百万输入 tokens $0.20、每百万输出 tokens $0.75，上线时优惠 80%。 来源：Inception blog（https://www.inceptionlabs.ai/blog/introducing-mercury-2-5）</description>
</item>
<item>
<title>OpenAI 正式提供 GPT-Rosalind</title>
<link>https://ai.gotry.io/zh/news/2026-09-08-openai-makes-gpt-rosalind-generally-available</link>
<guid isPermaLink="false">ai.gotry.io/news/2026-09-08-openai-makes-gpt-rosalind-generally-available</guid>
<pubDate>Tue, 08 Sep 2026 00:00:00 GMT</pubDate>
<category>发布</category>
<category>OpenAI</category>
<description>OpenAI 通过可信访问计划，向获批的内部生命科学研究开放 GPT-Rosalind（gpt-rosalind-research）。价格为每 100 万 input、缓存 input、output tokens 分别为 5、0.50、25 美元。计费自 2026 年 10 月 5 日开始。 来源：OpenAI API changelog（https://developers.openai.com/api/docs/changelog）</description>
</item>
<item>
<title>OpenAI 发布 GPT Image 2.5 Sunburst 与 Flare</title>
<link>https://ai.gotry.io/zh/news/2026-09-08-openai-releases-gpt-image-2-5-sunburst-and-flare</link>
<guid isPermaLink="false">ai.gotry.io/news/2026-09-08-openai-releases-gpt-image-2-5-sunburst-and-flare</guid>
<pubDate>Tue, 08 Sep 2026 00:00:00 GMT</pubDate>
<category>发布</category>
<category>OpenAI</category>
<description>OpenAI 发布 GPT Image 2.5 Sunburst 与 GPT Image 2.5 Flare，可通过 Image API 及 Responses 的图像生成工具进行生成与编辑。两者支持新的 xhigh 与 max 质量设置，并采用 GPT Image 2 的 token 费率。 来源：OpenAI API changelog（https://developers.openai.com/api/docs/changelog）</description>
</item>
<item>
<title>Kling AI 推出 Video Commerce API</title>
<link>https://ai.gotry.io/zh/news/2026-09-07-kuaishou-launches-video-commerce-api</link>
<guid isPermaLink="false">ai.gotry.io/news/2026-09-07-kuaishou-launches-video-commerce-api</guid>
<pubDate>Mon, 07 Sep 2026 00:00:00 GMT</pubDate>
<category>能力</category>
<category>Kuaishou Kling</category>
<description>Kling AI 通过 API 与 Agent 推出 Video Commerce。角色图加口播脚本可生成口播视频；加入商品图则生成推广镜头。可设置音频速度、宽高比、分辨率、音色、语速和背景音乐。 来源：Kling AI API updates（https://kling.ai/document-api/updates/api）</description>
</item>
<item>
<title>Ant Group 发布 Ling-3.0-flash-VL</title>
<link>https://ai.gotry.io/zh/news/2026-09-04-ant-group-releases-ling-3-0-flash-vl</link>
<guid isPermaLink="false">ai.gotry.io/news/2026-09-04-ant-group-releases-ling-3-0-flash-vl</guid>
<pubDate>Fri, 04 Sep 2026 00:00:00 GMT</pubDate>
<category>发布</category>
<category>Ant Group</category>
<description>2026年9月4日，Ant Group 推出多模态模型 Ling-3.0-flash-VL。用户可在聊天界面体验，并通过兼容 OpenAI 与 Anthropic 协议的 API 调用。 来源：Ant Ling changelog（https://developer.ant-ling.com/en/docs/getting-started/changelog/）</description>
</item>
<item>
<title>Google 发布 Lyria 3.5 音乐模型</title>
<link>https://ai.gotry.io/zh/news/2026-09-03-google-releases-lyria-3-5-music-model</link>
<guid isPermaLink="false">ai.gotry.io/news/2026-09-03-google-releases-lyria-3-5-music-model</guid>
<pubDate>Thu, 03 Sep 2026 00:00:00 GMT</pubDate>
<category>发布</category>
<category>Google</category>
<description>Lyria 3.5（lyria-3.5）正式可用，用于生成完整歌曲。它接受文本和图像输入，输出 44.1 kHz 立体声音频，并可控制时长与结构。 来源：Gemini API changelog（https://ai.google.dev/gemini-api/docs/changelog#09-03-2026）</description>
</item>
<item>
<title>OpenAI 为 GPT-6 Astra 增加长时间任务控制</title>
<link>https://ai.gotry.io/zh/news/2026-09-03-openai-adds-long-running-controls-for-gpt-6-astra</link>
<guid isPermaLink="false">ai.gotry.io/news/2026-09-03-openai-adds-long-running-controls-for-gpt-6-astra</guid>
<pubDate>Thu, 03 Sep 2026 00:00:00 GMT</pubDate>
<category>能力</category>
<category>OpenAI</category>
<description>Responses API 为 GPT-6 Astra 增加异步工具调用、通过 WebSocket 的回合中引导，以及对话中途调整推理力度，同时保留已缓存的提示前缀。 来源：OpenAI API changelog（https://developers.openai.com/api/docs/changelog）</description>
</item>
<item>
<title>OpenAI 发布 GPT-6 Astra</title>
<link>https://ai.gotry.io/zh/news/2026-09-03-openai-releases-gpt-6-astra</link>
<guid isPermaLink="false">ai.gotry.io/news/2026-09-03-openai-releases-gpt-6-astra</guid>
<pubDate>Thu, 03 Sep 2026 00:00:00 GMT</pubDate>
<category>发布</category>
<category>OpenAI</category>
<description>OpenAI 在 Responses 与 Chat Completions API 发布 GPT-6 Astra，用于推理、编程、计算机操作、研究与文档。它不支持 none 推理力度、自定义 temperature、top_p 和 logprobs。工具调用需使用 Responses API。 来源：OpenAI API changelog（https://developers.openai.com/api/docs/changelog）</description>
</item>
<item>
<title>Google 发布 Gemini 3.8 Flash</title>
<link>https://ai.gotry.io/zh/news/2026-09-02-google-releases-gemini-3-8-flash</link>
<guid isPermaLink="false">ai.gotry.io/news/2026-09-02-google-releases-gemini-3-8-flash</guid>
<pubDate>Wed, 02 Sep 2026 00:00:00 GMT</pubDate>
<category>发布</category>
<category>Google</category>
<description>gemini-3.8-flash 正式可用。Google 称其为最智能的 Flash 模型，面向长周期软件工程、智能体与企业工作流。 来源：Gemini API changelog（https://ai.google.dev/gemini-api/docs/changelog#09-02-2026）</description>
</item>
<item>
<title>Meta 发布 Muse Spark 1.3</title>
<link>https://ai.gotry.io/zh/news/2026-09-02-meta-releases-muse-spark-1-3</link>
<guid isPermaLink="false">ai.gotry.io/news/2026-09-02-meta-releases-muse-spark-1-3</guid>
<pubDate>Wed, 02 Sep 2026 00:00:00 GMT</pubDate>
<category>发布</category>
<category>Meta</category>
<description>2026 年 9 月 2 日，Meta 发布 Muse Spark 1.3，含 max reasoning 选项，已在 Muse Code 与 Meta Model API 提供。它面向更强的 agent 与编程任务，并更可靠地遵循长指令。 来源：Meta Superintelligence Labs research（https://research.meta.ai/blog/introducing-muse-spark-1-3）</description>
</item>
<item>
<title>Alibaba Qwen 发布 Qwen3.8 Max (2026-09-02)</title>
<link>https://ai.gotry.io/zh/news/2026-09-02-qwen3-8-max-0902-release</link>
<guid isPermaLink="false">ai.gotry.io/news/2026-09-02-qwen3-8-max-0902-release</guid>
<pubDate>Wed, 02 Sep 2026 00:00:00 GMT</pubDate>
<category>发布</category>
<category>Alibaba Qwen</category>
<description>上下文 1M tokens。Alibaba Cloud Model Studio：输入 ¥12.00、输出 ¥36.00 每百万 tokens。 来源：Alibaba Cloud Model Studio: newly released models（https://www.alibabacloud.com/help/en/model-studio/newly-released-models）</description>
</item>
<item>
<title>Anthropic 公布 Claude Fable 5.1 缓存读取价格</title>
<link>https://ai.gotry.io/zh/news/2026-09-01-anthropic-prices-claude-fable-5-1-cache-reads</link>
<guid isPermaLink="false">ai.gotry.io/news/2026-09-01-anthropic-prices-claude-fable-5-1-cache-reads</guid>
<pubDate>Tue, 01 Sep 2026 00:00:00 GMT</pubDate>
<category>价格</category>
<category>Anthropic</category>
<description>Claude Fable 5.1 与 Claude Mythos 5.1 的缓存读取为每百万 token 0.25 美元，即基础输入价的 0.025 倍，其他模型为 0.1 倍。缓存写入价格不变。 来源：Claude Developer Platform release notes（https://platform.claude.com/docs/en/release-notes/overview#september-1-2026）</description>
</item>
<item>
<title>Anthropic 发布 Claude Fable 5.1</title>
<link>https://ai.gotry.io/zh/news/2026-09-01-anthropic-releases-claude-fable-5-1</link>
<guid isPermaLink="false">ai.gotry.io/news/2026-09-01-anthropic-releases-claude-fable-5-1</guid>
<pubDate>Tue, 01 Sep 2026 00:00:00 GMT</pubDate>
<category>发布</category>
<category>Anthropic</category>
<description>Anthropic 推出 Claude Fable 5.1（claude-fable-5-1），作为 Claude Fable 5 的后继。默认上下文 100 万 token，最大输出 12.8 万，价格为每百万 token 10/50 美元，缓存读取 0.25 美元，已在 Claude API、Bedrock、AWS、Google Cloud 和 Microsoft Foundry 提供。 来源：Claude Developer Platform release notes（https://platform.claude.com/docs/en/release-notes/overview#september-1-2026）</description>
</item>
<item>
<title>Anthropic 发布 Claude Mythos 5.1</title>
<link>https://ai.gotry.io/zh/news/2026-09-01-claude-mythos-5-1-release</link>
<guid isPermaLink="false">ai.gotry.io/news/2026-09-01-claude-mythos-5-1-release</guid>
<pubDate>Tue, 01 Sep 2026 00:00:00 GMT</pubDate>
<category>发布</category>
<category>Anthropic</category>
<description>上下文 1M tokens。 来源：Claude Platform release notes（https://platform.claude.com/docs/en/release-notes/overview）</description>
</item>
<item>
<title>Google 为 Gemini 增加智能体式视频理解</title>
<link>https://ai.gotry.io/zh/news/2026-09-01-google-adds-agentic-video-understanding-to-gemini</link>
<guid isPermaLink="false">ai.gotry.io/news/2026-09-01-google-adds-agentic-video-understanding-to-gemini</guid>
<pubDate>Tue, 01 Sep 2026 00:00:00 GMT</pubDate>
<category>能力</category>
<category>Google</category>
<description>Gemini 3.7 Flash、3.6 Flash 与 3.5 Flash-Lite 在 Interactions 和 GenerateContent API 上提供智能体式视频理解。Google 称处理长视频时，token 用量可比静态处理最多减少 88%。 来源：Gemini API changelog（https://ai.google.dev/gemini-api/docs/changelog#09-01-2026）</description>
</item>
<item>
<title>Meta 发布 Muse Voice Transcribe</title>
<link>https://ai.gotry.io/zh/news/2026-09-01-meta-releases-muse-voice-transcribe</link>
<guid isPermaLink="false">ai.gotry.io/news/2026-09-01-meta-releases-muse-voice-transcribe</guid>
<pubDate>Tue, 01 Sep 2026 00:00:00 GMT</pubDate>
<category>发布</category>
<category>Meta</category>
<description>2026 年 9 月 1 日，Meta 发布实时音频模型 Muse Voice Transcribe，支持流式语音识别、20 位以上说话人分离、端点检测，以及多语言语码转换。 来源：Meta Superintelligence Labs research（https://research.meta.ai/blog/introducing-muse-voice-transcribe）</description>
</item>
<item>
<title>Alibaba Qwen 发布 qwen3.7-text-embedding-flash</title>
<link>https://ai.gotry.io/zh/news/2026-09-01-qwen3-7-text-embedding-flash-release</link>
<guid isPermaLink="false">ai.gotry.io/news/2026-09-01-qwen3-7-text-embedding-flash-release</guid>
<pubDate>Tue, 01 Sep 2026 00:00:00 GMT</pubDate>
<category>发布</category>
<category>Alibaba Qwen</category>
<description>qwen3.7-text-embedding-flash 已上线。 来源：模型上下架与更新-大模型服务平台百炼-阿里云_大模型服务平台百炼(Model Studio)-大模型服务平台百炼(Model Studio)-阿里云帮助中心（https://help.aliyun.com/zh/model-studio/newly-released-models）</description>
</item>
<item>
<title>Alibaba Qwen 发布 qwen3.7-text-rerank</title>
<link>https://ai.gotry.io/zh/news/2026-09-01-qwen3-7-text-rerank-release</link>
<guid isPermaLink="false">ai.gotry.io/news/2026-09-01-qwen3-7-text-rerank-release</guid>
<pubDate>Tue, 01 Sep 2026 00:00:00 GMT</pubDate>
<category>发布</category>
<category>Alibaba Qwen</category>
<description>qwen3.7-text-rerank 已上线。 来源：模型上下架与更新-大模型服务平台百炼-阿里云_大模型服务平台百炼(Model Studio)-大模型服务平台百炼(Model Studio)-阿里云帮助中心（https://help.aliyun.com/zh/model-studio/newly-released-models）</description>
</item>
<item>
<title>MongoDB 发布 rerank-3-lite</title>
<link>https://ai.gotry.io/zh/news/2026-09-01-rerank-3-lite-release</link>
<guid isPermaLink="false">ai.gotry.io/news/2026-09-01-rerank-3-lite-release</guid>
<pubDate>Tue, 01 Sep 2026 00:00:00 GMT</pubDate>
<category>发布</category>
<category>MongoDB</category>
<description>rerank-3-lite 已上线。 来源：Model Deprecations, Lifecycle States, and Support - Voyage AI by MongoDB - MongoDB Docs（https://www.mongodb.com/docs/voyageai/models/lifecycle/）</description>
</item>
<item>
<title>MongoDB 发布 rerank-3</title>
<link>https://ai.gotry.io/zh/news/2026-09-01-rerank-3-release</link>
<guid isPermaLink="false">ai.gotry.io/news/2026-09-01-rerank-3-release</guid>
<pubDate>Tue, 01 Sep 2026 00:00:00 GMT</pubDate>
<category>发布</category>
<category>MongoDB</category>
<description>rerank-3 已上线。 来源：Model Deprecations, Lifecycle States, and Support - Voyage AI by MongoDB - MongoDB Docs（https://www.mongodb.com/docs/voyageai/models/lifecycle/）</description>
</item>
<item>
<title>Mistral AI 正式提供 OCR 4.1</title>
<link>https://ai.gotry.io/zh/news/2026-08-31-mistral-makes-ocr-4-1-generally-available</link>
<guid isPermaLink="false">ai.gotry.io/news/2026-08-31-mistral-makes-ocr-4-1-generally-available</guid>
<pubDate>Mon, 31 Aug 2026 00:00:00 GMT</pubDate>
<category>发布</category>
<category>Mistral AI</category>
<description>2026年8月31日，Mistral AI 将 OCR 4.1（mistral-ocr-4-1）转为正式可用。 来源：Mistral AI changelog（https://docs.mistral.ai/resources/changelogs#date-2026-08-31）</description>
</item>
<item>
<title>Tencent 发布 Hy4 Preview</title>
<link>https://ai.gotry.io/zh/news/2026-08-28-hy4-preview-release</link>
<guid isPermaLink="false">ai.gotry.io/news/2026-08-28-hy4-preview-release</guid>
<pubDate>Fri, 28 Aug 2026 00:00:00 GMT</pubDate>
<category>发布</category>
<category>Tencent</category>
<description>上下文 1.02M tokens。Tencent Cloud TokenHub：输入 ¥6.00、输出 ¥18.00 每百万 tokens。 来源：Tencent: Tencent releases and open-sources Tencent Hy4 preview（https://www.tencent.com/tencent-releases-and-open-sources-tencent-hy4-preview/）</description>
</item>
<item>
<title>Alibaba Qwen 发布 qwen-mt-image-2.0</title>
<link>https://ai.gotry.io/zh/news/2026-08-28-qwen-mt-image-2-0-release</link>
<guid isPermaLink="false">ai.gotry.io/news/2026-08-28-qwen-mt-image-2-0-release</guid>
<pubDate>Fri, 28 Aug 2026 00:00:00 GMT</pubDate>
<category>发布</category>
<category>Alibaba Qwen</category>
<description>qwen-mt-image-2.0 已上线。 来源：Alibaba Cloud Model Studio: newly released models（https://www.alibabacloud.com/help/en/model-studio/newly-released-models）</description>
</item>
<item>
<title>Google 发布 Gemini Omni Flash</title>
<link>https://ai.gotry.io/zh/news/2026-08-27-gemini-omni-flash-release</link>
<guid isPermaLink="false">ai.gotry.io/news/2026-08-27-gemini-omni-flash-release</guid>
<pubDate>Thu, 27 Aug 2026 00:00:00 GMT</pubDate>
<category>发布</category>
<category>Google</category>
<description>Gemini Omni Flash 已上线。 来源：Gemini API deprecations（https://ai.google.dev/gemini-api/docs/deprecations?hl=en）</description>
</item>
<item>
<title>Google 发布 Gemini 3.5 Transcribe Live</title>
<link>https://ai.gotry.io/zh/news/2026-08-26-gemini-3-5-transcribe-live-release</link>
<guid isPermaLink="false">ai.gotry.io/news/2026-08-26-gemini-3-5-transcribe-live-release</guid>
<pubDate>Wed, 26 Aug 2026 00:00:00 GMT</pubDate>
<category>发布</category>
<category>Google</category>
<description>Gemini 3.5 Transcribe Live 已上线。 来源：Gemini API changelog（https://ai.google.dev/gemini-api/docs/changelog?hl=en）</description>
</item>
<item>
<title>Google 发布 Gemini 3.5 Transcribe</title>
<link>https://ai.gotry.io/zh/news/2026-08-26-gemini-3-5-transcribe-release</link>
<guid isPermaLink="false">ai.gotry.io/news/2026-08-26-gemini-3-5-transcribe-release</guid>
<pubDate>Wed, 26 Aug 2026 00:00:00 GMT</pubDate>
<category>发布</category>
<category>Google</category>
<description>Gemini 3.5 Transcribe 已上线。 来源：Gemini API changelog（https://ai.google.dev/gemini-api/docs/changelog?hl=en）</description>
</item>
<item>
<title>Zhipu AI 发布 GLM-5.3-Flash</title>
<link>https://ai.gotry.io/zh/news/2026-08-26-glm-5-3-flash-release</link>
<guid isPermaLink="false">ai.gotry.io/news/2026-08-26-glm-5-3-flash-release</guid>
<pubDate>Wed, 26 Aug 2026 00:00:00 GMT</pubDate>
<category>发布</category>
<category>Zhipu AI</category>
<description>上下文 1M tokens。Z.ai API：输入 $0.15、输出 $0.50 每百万 tokens。 来源：Z.ai new releases（https://docs.z.ai/release-notes/new-released）</description>
</item>
<item>
<title>OpenAI 弃用 whisper-1、gpt-4o-transcribe、gpt-4o-mini-transcribe、gpt-4o-transcribe-diarize</title>
<link>https://ai.gotry.io/zh/news/2026-08-26-openai-deprecated-whisper-1-and-3-more</link>
<guid isPermaLink="false">ai.gotry.io/news/2026-08-26-openai-deprecated-whisper-1-and-3-more</guid>
<pubDate>Wed, 26 Aug 2026 00:00:00 GMT</pubDate>
<category>下线</category>
<category>OpenAI</category>
<description>whisper-1、gpt-4o-transcribe、gpt-4o-mini-transcribe、gpt-4o-transcribe-diarize 已被弃用，将于 2027-02-26 下线。 来源：OpenAI API deprecations（https://developers.openai.com/api/docs/deprecations）</description>
</item>
<item>
<title>Alibaba Qwen 发布 Qwen3.8 Flash</title>
<link>https://ai.gotry.io/zh/news/2026-08-26-qwen3-8-flash-release</link>
<guid isPermaLink="false">ai.gotry.io/news/2026-08-26-qwen3-8-flash-release</guid>
<pubDate>Wed, 26 Aug 2026 00:00:00 GMT</pubDate>
<category>发布</category>
<category>Alibaba Qwen</category>
<description>上下文 1M tokens。Alibaba Cloud Model Studio：输入 $0.15、输出 $0.47 每百万 tokens。 来源：Alibaba Cloud Model Studio: newly released models（https://www.alibabacloud.com/help/en/model-studio/newly-released-models）</description>
</item>
<item>
<title>Alibaba Qwen 发布 Wan3.0 Video Prime</title>
<link>https://ai.gotry.io/zh/news/2026-08-20-wan3-0-video-prime-release</link>
<guid isPermaLink="false">ai.gotry.io/news/2026-08-20-wan3-0-video-prime-release</guid>
<pubDate>Thu, 20 Aug 2026 00:00:00 GMT</pubDate>
<category>发布</category>
<category>Alibaba Qwen</category>
<description>Wan3.0 Video Prime 已上线。 来源：Alibaba Cloud Model Studio: newly released models（https://www.alibabacloud.com/help/en/model-studio/newly-released-models）</description>
</item>
<item>
<title>Zhipu AI 发布 GLM-5.3</title>
<link>https://ai.gotry.io/zh/news/2026-08-18-glm-5-3-release</link>
<guid isPermaLink="false">ai.gotry.io/news/2026-08-18-glm-5-3-release</guid>
<pubDate>Tue, 18 Aug 2026 00:00:00 GMT</pubDate>
<category>发布</category>
<category>Zhipu AI</category>
<description>上下文 1M tokens。Z.ai API：输入 $1.40、输出 $4.40 每百万 tokens。 来源：Z.ai new releases（https://docs.z.ai/release-notes/new-released）</description>
</item>
<item>
<title>Alibaba Qwen 发布 qwen3.8-27b</title>
<link>https://ai.gotry.io/zh/news/2026-08-17-qwen3-8-27b-release</link>
<guid isPermaLink="false">ai.gotry.io/news/2026-08-17-qwen3-8-27b-release</guid>
<pubDate>Mon, 17 Aug 2026 00:00:00 GMT</pubDate>
<category>发布</category>
<category>Alibaba Qwen</category>
<description>上下文 1M tokens。Alibaba Cloud Model Studio：输入 $0.50、输出 $3.00 每百万 tokens。 来源：Alibaba Cloud Model Studio: newly released models（https://www.alibabacloud.com/help/en/model-studio/newly-released-models）</description>
</item>
<item>
<title>Google 发布 Gemini 3.7 Flash</title>
<link>https://ai.gotry.io/zh/news/2026-08-13-gemini-3-7-flash-release</link>
<guid isPermaLink="false">ai.gotry.io/news/2026-08-13-gemini-3-7-flash-release</guid>
<pubDate>Thu, 13 Aug 2026 00:00:00 GMT</pubDate>
<category>发布</category>
<category>Google</category>
<description>上下文 1.05M tokens。Gemini API：输入 $0.75、输出 $3.75 每百万 tokens。 来源：Gemini API changelog（https://ai.google.dev/gemini-api/docs/changelog?hl=en）</description>
</item>
<item>
<title>xAI 发布 Grok 4.6</title>
<link>https://ai.gotry.io/zh/news/2026-08-12-grok-4-6-release</link>
<guid isPermaLink="false">ai.gotry.io/news/2026-08-12-grok-4-6-release</guid>
<pubDate>Wed, 12 Aug 2026 00:00:00 GMT</pubDate>
<category>发布</category>
<category>xAI</category>
<description>上下文 500K tokens。xAI API：输入 $2.00、输出 $6.00 每百万 tokens。 来源：xAI release notes（https://docs.x.ai/developers/release-notes）</description>
</item>
</channel>
</rss>
