{"id":7166,"date":"2025-07-28T06:30:00","date_gmt":"2025-07-27T21:30:00","guid":{"rendered":"https:\/\/devneko.jp\/wordpress\/?p=7166"},"modified":"2025-07-27T13:32:39","modified_gmt":"2025-07-27T04:32:39","slug":"qwen3-coder","status":"publish","type":"post","link":"https:\/\/devneko.jp\/wordpress\/?p=7166","title":{"rendered":"Qwen3-Coder, Intern-S1, Step-Audio2, TeleChat2"},"content":{"rendered":"\n<p>Claude 4 sonnet\u30ec\u30d9\u30eb\u306eQwen3 Coder\uff08<a href=\"https:\/\/github.com\/QwenLM\/Qwen3-Coder\">QwenLM\/Qwen3-Coder: Qwen3-Coder is the code version of Qwen3, the large language model series developed by Qwen team, Alibaba Cloud.<\/a>\uff09\u3001235B MoE language model (Qwen3) \uff0b 6B Vision encoder (InternViT)\u3067\u5f37\u529b\u306a\u30de\u30eb\u30c1\u30e2\u30fc\u30c0\u30ebLRM Intern S1\uff08<a href=\"https:\/\/github.com\/InternLM\/Intern-S1\">InternLM\/Intern-S1<\/a>\uff09\u3001Kimi K2\u306e\u30c6\u30af\u30cb\u30ab\u30eb\u30ec\u30dd\u30fc\u30c8\u516c\u958b\uff08<a href=\"https:\/\/github.com\/MoonshotAI\/Kimi-K2\/blob\/main\/tech_report.pdf\">Kimi-K2\/tech_report.pdf at main \u00b7 MoonshotAI\/Kimi-K2<\/a>\uff09\u3001\u3068\u4e2d\u56fd\u306e\u30e2\u30c7\u30eb\u306b\u95a2\u3059\u308b\u8a71\u984c\u304c\u591a\u304b\u3063\u305f\u3002Qwen3-Instruct-2507\uff08<a href=\"https:\/\/github.com\/QwenLM\/Qwen3?tab=readme-ov-file#qwen3-instruct-2507\">QwenLM\/Qwen3: Qwen3 is the large language model series developed by Qwen team, Alibaba Cloud.<\/a>\uff09\u3067KIMI K2\u8d8a\u3048\u304c\u4e3b\u5f35\u3055\u308c\u305f\u308a\u3068\u7af6\u4e89\u304c\u6fc0\u3057\u3044\u3002<\/p>\n\n\n\n<p>\u97f3\u58f0\u95a2\u9023\u3067\u3082StepFun\u304b\u3089Step-Audio 2 Technical Report\u3001TeleAI\u304b\u3089TECHNICAL REPORT OF TELECHAT2, TELECHAT2.5 AND T1\u304c\u516c\u958b\u3055\u308c\u3066\u3044\u308b\u3002\u3044\u305a\u308c\u3082\u512a\u308c\u305f\u6027\u80fd\u3092\u4e3b\u5f35\u3002\u52a0\u3048\u3066GR-3\u306e\u3088\u3046\u306a\u30ed\u30dc\u30c3\u30c8\u95a2\u9023\u306e\u8ad6\u6587\u306b\u3082\u8208\u5473\u6d25\u3005\u3002<\/p>\n\n\n\n<p>\u305d\u3057\u3066\u3001\u3082\u3046\u9593\u3082\u306a\u304f\u3001GPT-5\u304c\u767a\u8868\u3055\u308c\u308b\u306f\u305a\u3067\u3001\u9032\u5316\u306f\u7d9a\u304d\u305d\u3046\u3002<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>Step-Audio 2 Technical Report&nbsp;<\/strong>[108.0]<br>Step-Audio 2\u306f\u3001\u696d\u754c\u306b\u304a\u3051\u308b\u97f3\u58f0\u7406\u89e3\u3068\u97f3\u58f0\u4f1a\u8a71\u306e\u305f\u3081\u306b\u8a2d\u8a08\u3055\u308c\u305f\u3001\u30a8\u30f3\u30c9\u30c4\u30fc\u30a8\u30f3\u30c9\u306e\u30de\u30eb\u30c1\u30e2\u30fc\u30c0\u30eb\u306a\u5927\u898f\u6a21\u8a00\u8a9e\u30e2\u30c7\u30eb\u3067\u3042\u308b\u3002 \u9045\u5ef6\u30aa\u30fc\u30c7\u30a3\u30aa\u30a8\u30f3\u30b3\u30fc\u30c0\u3068\u63a8\u8ad6\u4e2d\u5fc3\u5f37\u5316\u5b66\u7fd2(RL)\u3092\u7d71\u5408\u3059\u308b\u3053\u3068\u306b\u3088\u308a\u3001Step-Audio 2\u306f\u81ea\u52d5\u97f3\u58f0\u8a8d\u8b58(ASR)\u304a\u3088\u3073\u97f3\u58f0\u7406\u89e3\u306b\u304a\u3044\u3066\u6709\u671b\u306a\u6027\u80fd\u3092\u9054\u6210\u3059\u308b\u3002<br><a href=\"http:\/\/arxiv.org\/abs\/2507.16632v2\">\u8ad6\u6587<\/a>&nbsp;&nbsp;<a href=\"https:\/\/fugumt.com\/fugumt\/paper_check\/2507.16632v2\">\u53c2\u8003\u8a33\uff08\u30e1\u30bf\u30c7\u30fc\u30bf\uff09<\/a>&nbsp; &nbsp;(Thu, 24 Jul 2025 11:13:12 GMT)<\/li>\n\n\n\n<li>\u30ea\u30dd\u30b8\u30c8\u30ea\u306f<a href=\"https:\/\/github.com\/stepfun-ai\/Step-Audio2\">stepfun-ai\/Step-Audio2: Step-Audio 2 is an end-to-end multi-modal large language model designed for industry-strength audio understanding and speech conversation.<\/a><\/li>\n<\/ul>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>Technical Report of TeleChat2, TeleChat2.5 and T1&nbsp;<\/strong>[40.9]<br>\u6700\u65b0\u306eTeleChat\u30e2\u30c7\u30eb\u306b\u3064\u3044\u3066\u7d39\u4ecb\u3059\u308b: TeleChat2, TeleChat2.5, T1\u3002 \u30e2\u30c7\u30eb\u30a2\u30fc\u30ad\u30c6\u30af\u30c1\u30e3\u306e\u6700\u5c0f\u9650\u306e\u5909\u66f4\u306b\u3082\u304b\u304b\u308f\u3089\u305a\u3001\u65b0\u3057\u3044\u30b7\u30ea\u30fc\u30ba\u306f\u3001\u5f37\u5316\u3055\u308c\u305f\u30c8\u30ec\u30fc\u30cb\u30f3\u30b0\u6226\u7565\u306b\u3088\u3063\u3066\u3001\u5927\u5e45\u306a\u30d1\u30d5\u30a9\u30fc\u30de\u30f3\u30b9\u5411\u4e0a\u3092\u9054\u6210\u3059\u308b\u3002<br><a href=\"http:\/\/arxiv.org\/abs\/2507.18013v1\">\u8ad6\u6587<\/a>&nbsp;&nbsp;<a href=\"https:\/\/fugumt.com\/fugumt\/paper_check\/2507.18013v1\">\u53c2\u8003\u8a33\uff08\u30e1\u30bf\u30c7\u30fc\u30bf\uff09<\/a>&nbsp; &nbsp;(Thu, 24 Jul 2025 01:00:48 GMT)<\/li>\n\n\n\n<li>\u30ea\u30dd\u30b8\u30c8\u30ea\u306f<a href=\"https:\/\/github.com\/Tele-AI\/TeleChat2\">Tele-AI\/TeleChat2: \u661f\u8fb0\u8bed\u4e49\u5927\u6a21\u578bTeleChat2\u662f\u7531\u4e2d\u56fd\u7535\u4fe1\u4eba\u5de5\u667a\u80fd\u7814\u7a76\u9662\u7814\u53d1\u8bad\u7ec3\u7684\u5927\u8bed\u8a00\u6a21\u578b\uff0c\u662f\u9996\u4e2a\u5b8c\u5168\u56fd\u4ea7\u7b97\u529b\u8bad\u7ec3\u5e76\u5f00\u6e90\u7684\u5343\u4ebf\u53c2\u6570\u6a21\u578b<\/a><\/li>\n<\/ul>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>GR-3 Technical Report&nbsp;<\/strong>[21.9]<br>GR-3\u306f\u3001\u5927\u898f\u6a21\u306a\u8996\u899a\u8a00\u8a9e\u30a2\u30af\u30b7\u30e7\u30f3(VLA)\u30e2\u30c7\u30eb\u3067\u3042\u308b\u3002 \u62bd\u8c61\u6982\u5ff5\u3092\u542b\u3080\u65b0\u3057\u3044\u30aa\u30d6\u30b8\u30a7\u30af\u30c8\u3001\u74b0\u5883\u3001\u547d\u4ee4\u3092\u4e00\u822c\u5316\u3059\u308b\u969b\u3001\u4f8b\u5916\u7684\u306a\u80fd\u529b\u3092\u793a\u3059\u3002 GR-3\u306f\u3001\u4e21\u624b\u52d5\u64cd\u4f5c\u3084\u79fb\u52d5\u52d5\u4f5c\u3092\u5fc5\u8981\u3068\u3059\u308b\u30bf\u30b9\u30af\u3092\u542b\u3080\u3001\u9577\u3044\u6c34\u5e73\u304a\u3088\u3073\u5916\u63a5\u306a\u30bf\u30b9\u30af\u306e\u51e6\u7406\u306b\u9577\u3051\u3066\u3044\u308b\u3002<br><a href=\"http:\/\/arxiv.org\/abs\/2507.15493v1\">\u8ad6\u6587<\/a>&nbsp;&nbsp;<a href=\"https:\/\/fugumt.com\/fugumt\/paper_check\/2507.15493v1\">\u53c2\u8003\u8a33\uff08\u30e1\u30bf\u30c7\u30fc\u30bf\uff09<\/a>&nbsp; &nbsp;(Mon, 21 Jul 2025 10:54:13 GMT)<\/li>\n\n\n\n<li>\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u30b5\u30a4\u30c8\u306f<a href=\"https:\/\/seed.bytedance.com\/en\/GR3\">ByteDance Seed<\/a><\/li>\n<\/ul>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>Apple Intelligence Foundation Language Models: Tech Report 2025&nbsp;<\/strong>[246.0]<br>Apple\u306e\u30c7\u30d0\u30a4\u30b9\u3084\u30b5\u30fc\u30d3\u30b9\u306b\u307e\u305f\u304c\u3063\u3066Apple\u306e\u30a4\u30f3\u30c6\u30ea\u30b8\u30a7\u30f3\u30b9\u6a5f\u80fd\u3092\u99c6\u52d5\u3059\u308b2\u3064\u306e\u57fa\u790e\u8a00\u8a9e\u30e2\u30c7\u30eb\u3092\u7d39\u4ecb\u3057\u307e\u3059\u3002 \u3069\u3061\u3089\u306e\u30e2\u30c7\u30eb\u3082\u3001\u8cac\u4efb\u3042\u308bWeb\u30af\u30ed\u30fc\u30ea\u30f3\u30b0\u3092\u901a\u3058\u3066\u30bd\u30fc\u30b9\u3055\u308c\u308b\u5927\u898f\u6a21\u306a\u30de\u30eb\u30c1\u30ea\u30f3\u30ac\u30eb\u30c7\u30fc\u30bf\u30bb\u30c3\u30c8\u3068\u30de\u30eb\u30c1\u30e2\u30fc\u30c0\u30eb\u30c7\u30fc\u30bf\u30bb\u30c3\u30c8\u306b\u57fa\u3065\u3044\u3066\u30c8\u30ec\u30fc\u30cb\u30f3\u30b0\u3055\u308c\u3066\u3044\u308b\u3002 \u65b0\u3057\u3044Swift\u4e2d\u5fc3\u306eFoundation Models\u30d5\u30ec\u30fc\u30e0\u30ef\u30fc\u30af\u3067\u306f\u3001\u30ac\u30a4\u30c9\u4ed8\u304d\u751f\u6210\u3001\u5236\u7d04\u4ed8\u304d\u30c4\u30fc\u30eb\u547c\u3073\u51fa\u3057\u3001LoRA\u30a2\u30c0\u30d7\u30bf\u306e\u5fae\u8abf\u6574\u304c\u516c\u958b\u3055\u308c\u3066\u3044\u308b\u3002<br><a href=\"http:\/\/arxiv.org\/abs\/2507.13575v1\">\u8ad6\u6587<\/a>&nbsp;&nbsp;<a href=\"https:\/\/fugumt.com\/fugumt\/paper_check\/2507.13575v1\">\u53c2\u8003\u8a33\uff08\u30e1\u30bf\u30c7\u30fc\u30bf\uff09<\/a>&nbsp; &nbsp;(Thu, 17 Jul 2025 23:37:19 GMT)<\/li>\n\n\n\n<li>Apple Intelligence\u306e\u30c6\u30af\u30cb\u30ab\u30eb\u30ec\u30dd\u30fc\u30c8\u304carXiv\u306b\u516c\u958b\u3055\u308c\u3066\u3044\u305f\u3002<\/li>\n\n\n\n<li>\u300cWe found that AFM on-device model performs better than Qwen-2.5-3B, Gemma-3-4B and Gemma-3n-E4B on MMLU\/MMMLU, but it lags slightly behind Gemma-3n-E4B on MGSM. AFM on-device model performs lower than the larger Qwen-3-4B model.  AFM server models lag slightly to LLaMA 4 Scout, whose total size and active number of parameters are comparable, but has a bigger gap to larger models such as Qwen-3-235B and the proprietary GPT-4o.\u300d\u3068\u8a55\u4fa1\u3057\u3066\u3044\u308b\u3002<\/li>\n<\/ul>\n","protected":false},"excerpt":{"rendered":"<p>Claude 4 sonnet\u30ec\u30d9\u30eb\u306eQwen3 Coder\uff08QwenLM\/Qwen3-Coder: Qwen3-Coder is the code version of Qwen3, the large languag &hellip; <a href=\"https:\/\/devneko.jp\/wordpress\/?p=7166\" class=\"more-link\"><span class=\"screen-reader-text\">&#8220;Qwen3-Coder, Intern-S1, Step-Audio2, TeleChat2&#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":[223,232,524],"class_list":["post-7166","post","type-post","status-publish","format-standard","hentry","category-arxiv","tag-llm","tag-lrm","tag-524"],"_links":{"self":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/7166","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=7166"}],"version-history":[{"count":7,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/7166\/revisions"}],"predecessor-version":[{"id":7218,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/7166\/revisions\/7218"}],"wp:attachment":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=7166"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=7166"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=7166"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}