{"id":7386,"date":"2025-09-08T05:56:00","date_gmt":"2025-09-07T20:56:00","guid":{"rendered":"https:\/\/devneko.jp\/wordpress\/?p=7386"},"modified":"2025-09-07T12:27:23","modified_gmt":"2025-09-07T03:27:23","slug":"qwen3-max-k2-instruct-0905-longcat-flash-dream-coder-7b-kwai-keye-vl-1-5","status":"publish","type":"post","link":"https:\/\/devneko.jp\/wordpress\/?p=7386","title":{"rendered":"Qwen3-Max, K2-Instruct-0905, LongCat-Flash, Dream-Coder 7B, Kwai Keye-VL 1.5"},"content":{"rendered":"\n<p>\u5148\u9031\u3082LLM\/LRM\u754c\u9688\u306e\u30cb\u30e5\u30fc\u30b9\u306f\u591a\u304b\u3063\u305f\u3002Qwen3\u7cfb\u6700\u5927\u69cb\u6210\u306eQwen3 Max\u306e\u516c\u958b\uff08<a href=\"https:\/\/x.com\/alibaba_qwen\/status\/1963991502440562976\">X\u30e6\u30fc\u30b6\u30fc\u306eQwen\u3055\u3093: \u300cBig news: Introducing Qwen3-Max-Preview (Instruct) \u2014 our biggest model yet, with over 1 trillion parameters! \ud83d\ude80 Now available via Qwen Chat &amp; Alibaba Cloud API. Benchmarks show it beats our previous best, Qwen3-235B-A22B-2507. Internal tests + early user feedback confirm: https:\/\/t.co\/7vQTfHup1Z\u300d \/ X<\/a>\u3001<a href=\"https:\/\/www.alibabacloud.com\/help\/en\/model-studio\/models#161e7d77cfv1n\">Models and pricing &#8211; Alibaba Cloud Model Studio &#8211; Alibaba Cloud Documentation Center<\/a>\uff09\u3001Kimi K2\u306e\u30a2\u30c3\u30d7\u30c7\u30fc\u30c8\uff08<a href=\"https:\/\/x.com\/kimi_moonshot\/status\/1963802687230947698\">X\u30e6\u30fc\u30b6\u30fc\u306eKimi.ai\u3055\u3093: \u300cKimi K2-0905 update \ud83d\ude80 &#8211; Enhanced coding capabilities, esp. front-end &amp; tool-calling &#8211; Context length extended to 256k tokens &#8211; Improved integration with various agent scaffolds (e.g., Claude Code, Roo Code, etc) \ud83d\udd17 Weights &amp; code: https:\/\/t.co\/83sQekosr9 \ud83d\udcac Chat with new Kimi https:\/\/t.co\/mkOuBMwzpw\u300d \/ X<\/a>\u3001<a href=\"https:\/\/huggingface.co\/moonshotai\/Kimi-K2-Instruct-0905\">moonshotai\/Kimi-K2-Instruct-0905 \u00b7 Hugging Face<\/a>\uff09\u3084LongCat-Flash\u306e\u4ed6\u3001Dream-Coder 7B\u3001Kwai Keye-VL 1.5\u306a\u3069\u5c0f\u898f\u6a21\u3067\u3082\u30e6\u30cb\u30fc\u30af\u306a\u30e2\u30c7\u30eb\u3082\u767a\u8868\u3055\u308c\u3066\u3044\u308b\u3002<\/p>\n\n\n\n<p><a href=\"https:\/\/agentclientprotocol.com\/overview\/introduction\">Introduction &#8211; Agent Client Protocol<\/a>\uff08<a href=\"https:\/\/github.com\/zed-industries\/agent-client-protocol\">GitHub &#8211; zed-industries\/agent-client-protocol: A protocol for connecting any editor to any agent<\/a>\uff09\u3068\u3044\u3063\u305f\u30d7\u30ed\u30c8\u30b3\u30eb\u306e\u63d0\u6848\u306a\u3069\u5468\u8fba\u9818\u57df\u306b\u3082\u76ee\u304c\u96e2\u305b\u306a\u3044\u3002<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>LongCat-Flash Technical Report&nbsp;<\/strong>[165.7]<br>LongCat-Flash\u306f\u3001560\u30d3\u30ea\u30aa\u30f3\u30d1\u30e9\u30e1\u30fc\u30bf\u306eMixture-of-Experts (MoE)\u8a00\u8a9e\u30e2\u30c7\u30eb\u3067\u3042\u308b\u3002 \u8a08\u7b97\u52b9\u7387\u3068\u9ad8\u5ea6\u306a\u30a8\u30fc\u30b8\u30a7\u30f3\u30c8\u80fd\u529b\u306e\u4e21\u65b9\u306e\u305f\u3081\u306b\u8a2d\u8a08\u3055\u308c\u3066\u3044\u308b\u3002 30\u65e5\u4ee5\u5185\u306b20\u5146\u30c8\u30fc\u30af\u30f3\u4ee5\u4e0a\u306e\u30e2\u30c7\u30eb\u30c8\u30ec\u30fc\u30cb\u30f3\u30b0\u3092\u5b8c\u4e86\u3057\u3001100\u30c8\u30fc\u30af\u30f3\/\u79d2 (TPS) \u4ee5\u4e0a\u306e\u63a8\u8ad6\u30920.70\u30d1\u30fc\u30bb\u30f3\u30c8\u306e\u30a2\u30a6\u30c8\u30d7\u30c3\u30c8\u30c8\u30fc\u30af\u30f3\u3067\u9054\u6210\u3057\u307e\u3057\u305f\u3002<br><a href=\"http:\/\/arxiv.org\/abs\/2509.01322v1\">\u8ad6\u6587<\/a>&nbsp;&nbsp;<a href=\"https:\/\/fugumt.com\/fugumt\/paper_check\/2509.01322v1\">\u53c2\u8003\u8a33\uff08\u30e1\u30bf\u30c7\u30fc\u30bf\uff09<\/a>&nbsp; &nbsp;(Mon, 01 Sep 2025 10:05:45 GMT)<\/li>\n\n\n\n<li>560B MoE\u69cb\u6210\u3001\u300cAs a non-thinking model, LongCat-Flash achieves performance comparable to state-of-the-art non-thinking models, including DeepSeek-V3.1 [DeepSeek-AI et al , 2025] and Kimi-K2 [Team et al , 2025], while using fewer parameters and offering faster inference speed. Specifically, LongCat-Flash scores 86.5 on ArenaHard-V2, 39.5 on TerminalBench, and 67.7 on \u03c4 2-Bench, demonstrating robust capabilities in general domains, coding, and agentic tool use.\u300d<\/li>\n\n\n\n<li>\u30ea\u30dd\u30b8\u30c8\u30ea\u306f<a href=\"https:\/\/github.com\/meituan-longcat\/LongCat-Flash-Chat\">GitHub &#8211; meituan-longcat\/LongCat-Flash-Chat<\/a><\/li>\n<\/ul>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>Dream-Coder 7B: An Open Diffusion Language Model for Code&nbsp;<\/strong>[99.1]<br>\u305d\u3053\u3067,Dream-Coder 7B\u3092\u63d0\u6848\u3059\u308b\u3002Dream-Coder 7B\u306f,\u4efb\u610f\u306e\u9806\u5e8f\u751f\u6210\u80fd\u529b\u3092\u793a\u3059\u30b3\u30fc\u30c9\u751f\u6210\u306e\u305f\u3081\u306e,\u30aa\u30fc\u30d7\u30f3\u30bd\u30fc\u30b9\u306e\u96e2\u6563\u62e1\u6563\u8a00\u8a9e\u30e2\u30c7\u30eb\u3067\u3042\u308b\u3002 \u53b3\u5bc6\u306b\u5de6\u304b\u3089\u53f3\u306b\u30c7\u30b3\u30fc\u30c9\u3059\u308b\u5f93\u6765\u306e\u81ea\u5df1\u56de\u5e30(AR)\u30e2\u30c7\u30eb\u3068\u306f\u7570\u306a\u308a\u3001\u30c9\u30ea\u30fc\u30e0\u30b3\u30fc\u30c07B\u306f\u30b3\u30fc\u30c7\u30a3\u30f3\u30b0\u30bf\u30b9\u30af\u306b\u57fa\u3065\u3044\u3066\u30c7\u30b3\u30fc\u30c9\u6226\u7565\u3092\u9069\u5fdc\u7684\u306b\u6c7a\u5b9a\u3059\u308b\u3002<br><a href=\"http:\/\/arxiv.org\/abs\/2509.01142v1\">\u8ad6\u6587<\/a>&nbsp;&nbsp;<a href=\"https:\/\/fugumt.com\/fugumt\/paper_check\/2509.01142v1\">\u53c2\u8003\u8a33\uff08\u30e1\u30bf\u30c7\u30fc\u30bf\uff09<\/a>&nbsp; &nbsp;(Mon, 01 Sep 2025 05:30:56 GMT)<\/li>\n\n\n\n<li>\u30b3\u30fc\u30c7\u30a3\u30f3\u30b0\u30bf\u30b9\u30af\u5f37\u5316\u306e\u62e1\u6563\u30e2\u30c7\u30eb<\/li>\n\n\n\n<li>\u30ea\u30dd\u30b8\u30c8\u30ea\u306f<a href=\"https:\/\/github.com\/DreamLM\/Dream-Coder\">GitHub &#8211; DreamLM\/Dream-Coder<\/a><\/li>\n<\/ul>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>Kwai Keye-VL 1.5 Technical Report&nbsp;<\/strong>[91.3]<br>\u672c\u7a3f\u3067\u306f\u3001\u30d3\u30c7\u30aa\u7406\u89e3\u306b\u304a\u3051\u308b\u6839\u672c\u7684\u306a\u8ab2\u984c\u30923\u3064\u306e\u91cd\u8981\u306a\u30a4\u30ce\u30d9\u30fc\u30b7\u30e7\u30f3\u3092\u901a\u3058\u3066\u89e3\u6c7a\u3059\u308bKeye-VL-1.5\u3092\u7d39\u4ecb\u3059\u308b\u3002 \u307e\u305a,\u30d5\u30ec\u30fc\u30e0\u9593\u306e\u985e\u4f3c\u6027\u306b\u57fa\u3065\u3044\u3066\u52d5\u7684\u306b\u8a08\u7b97\u8cc7\u6e90\u3092\u5272\u308a\u5f53\u3066\u308bSlow-Fast\u30d3\u30c7\u30aa\u7b26\u53f7\u5316\u65b9\u5f0f\u3092\u63d0\u6848\u3059\u308b\u3002 \u6b21\u306b,\u30e2\u30c7\u30eb\u306e\u30b3\u30f3\u30c6\u30ad\u30b9\u30c8\u9577\u30928K\u304b\u3089128K\u307e\u3067\u4f53\u7cfb\u7684\u306b\u62e1\u5f35\u3059\u308b4\u6bb5\u968e\u4e8b\u524d\u5b66\u7fd2\u624b\u6cd5\u3092\u63d0\u6848\u3059\u308b\u3002 \u7b2c3\u306b\u3001\u63a8\u8ad6\u306e\u5f37\u5316\u3068\u4eba\u9593\u306e\u55dc\u597d\u306e\u6574\u5408\u6027\u306b\u7126\u70b9\u3092\u5f53\u3066\u305f\u7dcf\u5408\u7684\u306a\u5f8c\u5b66\u7fd2\u30d1\u30a4\u30d7\u30e9\u30a4\u30f3\u3092\u958b\u767a\u3059\u308b\u3002<br><a href=\"http:\/\/arxiv.org\/abs\/2509.01563v1\">\u8ad6\u6587<\/a>&nbsp;&nbsp;<a href=\"https:\/\/fugumt.com\/fugumt\/paper_check\/2509.01563v1\">\u53c2\u8003\u8a33\uff08\u30e1\u30bf\u30c7\u30fc\u30bf\uff09<\/a>&nbsp; &nbsp;(Mon, 01 Sep 2025 15:46:58 GMT)<\/li>\n\n\n\n<li>\u300cKeye-VL-1.5-8B establishes new state-of-the-art performance among models of similar scale, demonstrating superior results on video-centric benchmarks while maintaining competitive performance on general multimodal and reasoning tasks.\u300d\u3068\u30d3\u30c7\u30aa\u3092\u6271\u3048\u308b\u30e2\u30c7\u30eb<\/li>\n\n\n\n<li>\u30ea\u30dd\u30b8\u30c8\u30ea\u306f<a href=\"https:\/\/github.com\/Kwai-Keye\/Keye\">GitHub &#8211; Kwai-Keye\/Keye<\/a><\/li>\n<\/ul>\n","protected":false},"excerpt":{"rendered":"<p>\u5148\u9031\u3082LLM\/LRM\u754c\u9688\u306e\u30cb\u30e5\u30fc\u30b9\u306f\u591a\u304b\u3063\u305f\u3002Qwen3\u7cfb\u6700\u5927\u69cb\u6210\u306eQwen3 Max\u306e\u516c\u958b\uff08X\u30e6\u30fc\u30b6\u30fc\u306eQwen\u3055\u3093: \u300cBig news: Introducing Qwen3-Max-Preview (Instruc &hellip; <a href=\"https:\/\/devneko.jp\/wordpress\/?p=7386\" class=\"more-link\"><span class=\"screen-reader-text\">&#8220;Qwen3-Max, K2-Instruct-0905, LongCat-Flash, Dream-Coder 7B, Kwai Keye-VL 1.5&#8221; \u306e<\/span>\u7d9a\u304d\u3092\u8aad\u3080<\/a><\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[2],"tags":[114,223,232],"class_list":["post-7386","post","type-post","status-publish","format-standard","hentry","category-arxiv","tag-diffusion-model","tag-llm","tag-lrm"],"_links":{"self":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/7386","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=7386"}],"version-history":[{"count":3,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/7386\/revisions"}],"predecessor-version":[{"id":7416,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/7386\/revisions\/7416"}],"wp:attachment":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=7386"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=7386"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=7386"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}