{"id":272,"date":"2026-07-08T11:30:23","date_gmt":"2026-07-08T02:30:23","guid":{"rendered":"https:\/\/www.wetch.co.jp\/ai\/?p=272"},"modified":"2026-07-08T11:30:23","modified_gmt":"2026-07-08T02:30:23","slug":"kimi-k2-7-code%e3%83%bbmistral-ocr-4%e3%83%bbdeepseek-v4-flash-2026%e5%b9%b46%e6%9c%88%e3%82%aa%e3%83%bc%e3%83%97%e3%83%b3%e3%82%a6%e3%82%a7%e3%82%a4%e3%83%88llm%e3%82%b3%e3%83%bc","status":"publish","type":"post","link":"https:\/\/www.wetch.co.jp\/ai\/2026\/07\/08\/kimi-k2-7-code%e3%83%bbmistral-ocr-4%e3%83%bbdeepseek-v4-flash-2026%e5%b9%b46%e6%9c%88%e3%82%aa%e3%83%bc%e3%83%97%e3%83%b3%e3%82%a6%e3%82%a7%e3%82%a4%e3%83%88llm%e3%82%b3%e3%83%bc\/","title":{"rendered":"Kimi K2.7 Code\u30fbMistral OCR 4\u30fbDeepSeek V4-Flash\u2014\u20142026\u5e746\u6708\u30aa\u30fc\u30d7\u30f3\u30a6\u30a7\u30a4\u30c8LLM\u30b3\u30fc\u30c7\u30a3\u30f3\u30b0\u6700\u524d\u7dda"},"content":{"rendered":"<p><em>2026\u5e746\u6708\u3001\u30aa\u30fc\u30d7\u30f3\u30a6\u30a7\u30a4\u30c8\uff08\u516c\u958b\u91cd\u307f\uff09LLM\u306e\u30b3\u30fc\u30c7\u30a3\u30f3\u30b0\u6027\u80fd\u304c\u30d7\u30ed\u30d7\u30e9\u30a4\u30a8\u30bf\u30ea\u30e2\u30c7\u30eb\u306b\u8089\u8584\u3059\u308b\u52d5\u304d\u304c\u52a0\u901f\u3057\u3066\u3044\u308b\u3002Moonshot AI\u306e<strong>Kimi K2.7 Code<\/strong>\u304cSWE-bench Pro\u3067\u4e3b\u8981\u5546\u7528\u30e2\u30c7\u30eb\u3092\u4e0a\u56de\u308a\u3001Mistral\u306f\u30bb\u30eb\u30d5\u30db\u30b9\u30c8OCR\u30a8\u30f3\u30b8\u30f3<strong>OCR 4<\/strong>\u3092\u30aa\u30fc\u30d7\u30f3\u30bd\u30fc\u30b9\u3067\u6295\u5165\u3001Alibaba\u3068DeepSeek\u306f\u9ad8\u30b3\u30b9\u30d1\u306a\u30aa\u30fc\u30d7\u30f3\u30a6\u30a7\u30a4\u30c8MoE\u30e2\u30c7\u30eb\u3092\u76f8\u6b21\u3044\u3067\u30a2\u30c3\u30d7\u30c7\u30fc\u30c8\u3057\u305f\u3002\u672c\u7a3f\u3067\u306f\u4eca\u9031\u306e\u4e3b\u8981\u30c8\u30d4\u30c3\u30af\u3092\u6a2a\u65ad\u7684\u306b\u307e\u3068\u3081\u308b\u3002<\/em><\/p>\n<figure class=\"wp-block-image size-large\"><img decoding=\"async\" src=\"https:\/\/www.wetch.co.jp\/ai\/wp-content\/uploads\/2026\/06\/ill-oss.png\" alt=\"\u30aa\u30fc\u30d7\u30f3\u30bd\u30fc\u30b9LLM\u30a4\u30e9\u30b9\u30c8\"\/><\/figure>\n<h2>Kimi K2.7 Code\u2014\u201412\u6642\u9593\u81ea\u5f8b\u30b3\u30fc\u30c7\u30a3\u30f3\u30b0\u3068300\u30a8\u30fc\u30b8\u30a7\u30f3\u30c8\u30b9\u30a6\u30a9\u30fc\u30e0<\/h2>\n<p><a href=\"https:\/\/platform.moonshot.ai\" target=\"_blank\" rel=\"noopener\">Moonshot AI<\/a> \u306f2026\u5e746\u670812\u65e5\u3001\u30aa\u30fc\u30d7\u30f3\u30a6\u30a7\u30a4\u30c8\u306e\u30b3\u30fc\u30c7\u30a3\u30f3\u30b0\u7279\u5316\u30e2\u30c7\u30eb <a href=\"https:\/\/kimi.com\/code\" target=\"_blank\" rel=\"noopener\"><strong>Kimi K2.7 Code<\/strong><\/a> \u3092\u30ea\u30ea\u30fc\u30b9\u3057\u305f\u3002\u524d\u4e16\u4ee3\u306e K2.6 \u6bd4\u3067\u30b3\u30fc\u30c7\u30a3\u30f3\u30b0\u30d9\u30f3\u30c1\u30de\u30fc\u30af\uff08Kimi Code Bench v2\uff09\u304c <strong>+21.8%<\/strong>\u3001Program Bench \u304c +11.0%\u3001MLS Bench Lite \u304c +31.5% \u5411\u4e0a\u3057\u3066\u3044\u308b\u3002\u63a8\u8ad6\u30c8\u30fc\u30af\u30f3\u4f7f\u7528\u91cf\u306f30%\u524a\u6e1b\u3055\u308c\u3066\u304a\u308a\u3001\u300c\u8003\u3048\u3059\u304e\uff08overthinking\uff09\u300d\u554f\u984c\u3092\u5927\u5e45\u306b\u6539\u5584\u3057\u305f\u3002<\/p>\n<ul>\n<li><strong>\u30e2\u30c7\u30eb\u898f\u6a21<\/strong>: \u7dcf\u30d1\u30e9\u30e1\u30fc\u30bf\u6570 1T\uff08MoE\uff09\u3001\u30a2\u30af\u30c6\u30a3\u30d6\u30d1\u30e9\u30e1\u30fc\u30bf 32B<\/li>\n<li><strong>\u30b3\u30f3\u30c6\u30ad\u30b9\u30c8<\/strong>: 256K \u30c8\u30fc\u30af\u30f3<\/li>\n<li><strong>\u30d9\u30f3\u30c1\u30de\u30fc\u30af<\/strong>: SWE-bench Pro 58.6 \/ SWE-bench Verified 80.2 \/ AIME 2026 96.4<\/li>\n<li><strong>\u81ea\u5f8b\u30a8\u30fc\u30b8\u30a7\u30f3\u30c8<\/strong>: 12\u6642\u9593\u4ee5\u4e0a\u306e\u7121\u4eba\u30b3\u30fc\u30c7\u30a3\u30f3\u30b0\u30bb\u30c3\u30b7\u30e7\u30f3\u3001\u6700\u5927300\u30b5\u30d6\u30a8\u30fc\u30b8\u30a7\u30f3\u30c8\u30fb4,000\u30c4\u30fc\u30eb\u30b3\u30fc\u30eb\u3092\u78ba\u8a8d<\/li>\n<li><strong>\u30de\u30eb\u30c1\u30e2\u30fc\u30c0\u30eb<\/strong>: \u30cd\u30a4\u30c6\u30a3\u30d6\u753b\u50cf\u30fb\u52d5\u753b\u5165\u529b\u5bfe\u5fdc\uff08MoonViT \u30d3\u30b8\u30e7\u30f3\u30a8\u30f3\u30b3\u30fc\u30c0\u642d\u8f09\uff09<\/li>\n<li><strong>\u30e9\u30a4\u30bb\u30f3\u30b9<\/strong>: Modified MIT\uff08\u5546\u7528\u5229\u7528\u53ef\uff09<\/li>\n<\/ul>\n<p>\u5e7b\u899a\uff08\u30cf\u30eb\u30b7\u30cd\u30fc\u30b7\u30e7\u30f3\uff09\u7387\u3082 K2.6 \u306e\u7d0465%\u304b\u3089 <strong>39%<\/strong> \u306b\u4f4e\u4e0b\u3002\u9577\u6642\u9593\u306e\u81ea\u5f8b\u7684\u306a\u30bd\u30d5\u30c8\u30a6\u30a7\u30a2\u958b\u767a\u30bf\u30b9\u30af\u3092\u5b89\u5b9a\u3057\u3066\u3053\u306a\u305b\u308b\u70b9\u304c\u969b\u7acb\u3064\u3002\u30e2\u30c7\u30eb\u306f <a href=\"https:\/\/platform.moonshot.ai\" target=\"_blank\" rel=\"noopener\">Kimi API<\/a> \u304a\u3088\u3073 <a href=\"https:\/\/kimi.com\/code\" target=\"_blank\" rel=\"noopener\">Kimi Code<\/a> \u304b\u3089\u5229\u7528\u3067\u304d\u3001Hugging Face \u3067\u3082\u516c\u958b\u3055\u308c\u3066\u3044\u308b\u3002<\/p>\n<p>\u51fa\u5178: <a href=\"https:\/\/x.com\/Kimi_Moonshot\/status\/2065377579130142937\" target=\"_blank\" rel=\"noopener\">Moonshot AI \u516c\u5f0f X (2026\u5e746\u670812\u65e5)<\/a> \/ <a href=\"https:\/\/www.vm0.ai\/en\/models\/kimi-k2-7-code\" target=\"_blank\" rel=\"noopener\">VM0 \u30e2\u30c7\u30eb\u8a73\u7d30<\/a><\/p>\n<h2>Mistral OCR 4\u2014\u2014170\u8a00\u8a9e\u5bfe\u5fdc\u30fb\u30bb\u30eb\u30d5\u30db\u30b9\u30c8\u3067\u52d5\u304f\u6587\u66f8\u30a4\u30f3\u30c6\u30ea\u30b8\u30a7\u30f3\u30b9<\/h2>\n<p><a href=\"https:\/\/mistral.ai\" target=\"_blank\" rel=\"noopener\">Mistral AI<\/a> \u306f2026\u5e746\u670823\u65e5\u306b <a href=\"https:\/\/mistral.ai\/news\/ocr-4\" target=\"_blank\" rel=\"noopener\"><strong>Mistral OCR 4<\/strong><\/a> \u3092\u516c\u958b\u3057\u305f\u3002\u4f01\u696d\u306e\u6587\u66f8\u51e6\u7406\u30d1\u30a4\u30d7\u30e9\u30a4\u30f3\u3092\u5ff5\u982d\u306b\u8a2d\u8a08\u3055\u308c\u305f\u30e2\u30c7\u30eb\u3067\u3001<strong>170\u8a00\u8a9e\u30fb10\u8a00\u8a9e\u30b0\u30eb\u30fc\u30d7<\/strong>\u306b\u5bfe\u5fdc\u3057\u3001\u30c6\u30ad\u30b9\u30c8\u62bd\u51fa\u3068\u540c\u6642\u306b\u30d0\u30a6\u30f3\u30c7\u30a3\u30f3\u30b0\u30dc\u30c3\u30af\u30b9\u30fb\u30d6\u30ed\u30c3\u30af\u5206\u985e\u30fb\u30a4\u30f3\u30e9\u30a4\u30f3\u4fe1\u983c\u30b9\u30b3\u30a2\u3092\u51fa\u529b\u3059\u308b\u3002<\/p>\n<ul>\n<li><strong>\u30c7\u30d7\u30ed\u30a4\u5f62\u614b<\/strong>: \u5358\u4e00\u30b3\u30f3\u30c6\u30ca\u3067\u5b8c\u7d50\u3059\u308b\u30bb\u30eb\u30d5\u30db\u30b9\u30c8\u5bfe\u5fdc\uff08\u30c7\u30fc\u30bf\u4e3b\u6a29\u30fb\u30b3\u30f3\u30d7\u30e9\u30a4\u30a2\u30f3\u30b9\u8981\u4ef6\u3092\u6e80\u305f\u3057\u3084\u3059\u3044\uff09<\/li>\n<li><strong>\u9ad8\u30b9\u30eb\u30fc\u30d7\u30c3\u30c8<\/strong>: \u30b3\u30b9\u30c8\u52b9\u7387\u306e\u9ad8\u3044\u30d0\u30c3\u30c1\u51e6\u7406\u306b\u6700\u9069\u5316<\/li>\n<li><strong>Mistral Search Toolkit \u7d71\u5408<\/strong>: AI Now Summit 2026 \u3067\u767a\u8868\u3055\u308c\u305f\u540c\u793e\u30aa\u30fc\u30d7\u30f3\u30bd\u30fc\u30b9\u691c\u7d22\u30d5\u30ec\u30fc\u30e0\u30ef\u30fc\u30af\u306e\u30a4\u30f3\u30b8\u30a7\u30b9\u30b7\u30e7\u30f3 \u30b3\u30f3\u30dd\u30fc\u30cd\u30f3\u30c8\u3068\u3057\u3066\u6a5f\u80fd\uff08RAG\u30fb\u30a8\u30f3\u30bf\u30fc\u30d7\u30e9\u30a4\u30ba\u691c\u7d22\u30fb\u30c9\u30e1\u30a4\u30f3\u7279\u5316\u578b\u691c\u7d22\u30d1\u30a4\u30d7\u30e9\u30a4\u30f3\u306b\u7d44\u307f\u8fbc\u307f\u53ef\u80fd\uff09<\/li>\n<\/ul>\n<p>\u307e\u305f\u3001Mistral \u306f\u7c73\u56fd\u304c <a href=\"https:\/\/deccanfounders.com\/2026\/18\/news\/mistral-promotes-open-source-ai-after-us-restricts-anthropic-model-access\/\" target=\"_blank\" rel=\"noopener\">\u4e00\u90e8\u306e\u5546\u7528\u30e2\u30c7\u30eb\u3078\u306e\u30a2\u30af\u30bb\u30b9\u3092\u5236\u9650<\/a>\u3059\u308b\u52d5\u304d\u304c\u51fa\u305f\u76f4\u5f8c\u304b\u3089\u3001\u30aa\u30fc\u30d7\u30f3\u30bd\u30fc\u30b9\u6226\u7565\u3092\u524d\u9762\u306b\u62bc\u3057\u51fa\u3057\u3066\u304a\u308a\u3001\u6b27\u5dde\u30fb\u30a2\u30b8\u30a2\u5e02\u5834\u3067\u306e\u300c\u4e3b\u6a29AI\u300d\u4ee3\u66ff\u3068\u3057\u3066\u5b58\u5728\u611f\u3092\u9ad8\u3081\u3066\u3044\u308b\u3002\u65b0\u305f\u306a\u30b3\u30fc\u30c7\u30a3\u30f3\u30b0\u30a8\u30fc\u30b8\u30a7\u30f3\u30c8\u7279\u5316\u30e2\u30c7\u30eb <strong>Devstral 2<\/strong> \u3082\u3053\u306e\u6d41\u308c\u3067\u4eca\u6708\u6295\u5165\u3055\u308c\u3066\u304a\u308a\u300124B \u30d1\u30e9\u30e1\u30fc\u30bf\u306a\u304c\u3089\u30a8\u30fc\u30b8\u30a7\u30f3\u30c8\u30ef\u30fc\u30af\u30d5\u30ed\u30fc\u3067\u306e\u512a\u308c\u305f\u5b9f\u7e3e\u304c\u5831\u544a\u3055\u308c\u3066\u3044\u308b\u3002<\/p>\n<figure class=\"wp-block-image size-large\"><img decoding=\"async\" src=\"https:\/\/www.wetch.co.jp\/ai\/wp-content\/uploads\/2026\/06\/code-screen-cc0.jpg\" alt=\"\u30b3\u30fc\u30c9\u304c\u6620\u308b\u30b9\u30af\u30ea\u30fc\u30f3\"\/><figcaption><small>Photo: <a href=\"https:\/\/www.rawpixel.com\/image\/3301681\/free-photo-image-data-coding-computer-business\" target=\"_blank\" rel=\"noopener\">rawpixel.com<\/a> \/ CC0\uff08Openverse\u7d4c\u7531\uff09<\/small><\/figcaption><\/figure>\n<p>\u51fa\u5178: <a href=\"https:\/\/mistral.ai\/news\/ocr-4\" target=\"_blank\" rel=\"noopener\">Mistral AI \u516c\u5f0f\u30d6\u30ed\u30b0\u300cMistral OCR 4\u300d<\/a><\/p>\n<h2>Qwen3-Coder \u3068 DeepSeek V4-Flash\u2014\u2014\u5546\u7528\u30e9\u30a4\u30bb\u30f3\u30b9\u00d7\u9ad8\u30b3\u30b9\u30d1\u306e\u4e8c\u5927\u9078\u629e\u80a2<\/h2>\n<p>Alibaba \u3068 DeepSeek \u306f\u3001\u4eca\u6708\u3082\u6d88\u8cbb\u8005\u30cf\u30fc\u30c9\u30a6\u30a7\u30a2\u3067\u5b9f\u7528\u306b\u306a\u308b\u5927\u898f\u6a21\u30e2\u30c7\u30eb\u3092\u30ea\u30ea\u30fc\u30b9\u3057\u7d9a\u3051\u3066\u3044\u308b\u3002<\/p>\n<h3><a href=\"https:\/\/qwenlm.github.io\/\" target=\"_blank\" rel=\"noopener\">Qwen3-Coder 480B-A35B<\/a>\uff08Alibaba \/ Apache-2.0\uff09<\/h3>\n<p>SWE-bench Verified \u30b9\u30b3\u30a2 <strong>69.6%<\/strong> \u306f\u30aa\u30fc\u30d7\u30f3\u30a6\u30a7\u30a4\u30c8\u30e2\u30c7\u30eb\u306e\u4e2d\u3067\u73fe\u72b6\u6700\u9ad8\u6c34\u6e96\uff08\u30b7\u30f3\u30b0\u30eb\u8a66\u884c\uff09\u3002MoE \u69cb\u9020\u306b\u3088\u308a 480B \u7dcf\u30d1\u30e9\u30e1\u30fc\u30bf\u306e\u3046\u3061\u30a2\u30af\u30c6\u30a3\u30d6\u306f 35B \u306b\u6291\u3048\u3089\u308c\u3001\u30b5\u30fc\u30d3\u30f3\u30b0\u52b9\u7387\u304c\u9ad8\u3044\u3002\u307e\u305f\u30ed\u30fc\u30ab\u30eb\u5411\u3051\u306b\u306f <strong>Qwen3.6 27B<\/strong>\uff0877.2% SWE-bench\uff09\u304c Q4 \u91cf\u5b50\u5316\u3067\u7d0424GB\u306b\u53ce\u307e\u308a\u3001\u30b3\u30f3\u30b7\u30e5\u30fc\u30de GPU \u74b0\u5883\u3067\u3082\u52d5\u4f5c\u3059\u308b\u3002<\/p>\n<h3><a href=\"https:\/\/www.deepseek.com\/\" target=\"_blank\" rel=\"noopener\">DeepSeek V4-Flash<\/a>\uff08DeepSeek \/ MIT\uff09<\/h3>\n<p>284B \u7dcf\u30d1\u30e9\u30e1\u30fc\u30bf\u30fb13B \u30a2\u30af\u30c6\u30a3\u30d6\u306e MoE \u30e2\u30c7\u30eb\u3067\u3001<strong>1M\u30c8\u30fc\u30af\u30f3<\/strong>\u30b3\u30f3\u30c6\u30ad\u30b9\u30c8\u3092\u6301\u3061\u306a\u304c\u3089\u5358\u4e00 H100 GPU \u3067\u7a3c\u50cd\u3067\u304d\u308b\u3002MIT \u30e9\u30a4\u30bb\u30f3\u30b9\u306e\u305f\u3081\u5546\u7528 SaaS \u3078\u306e\u7d44\u307f\u8fbc\u307f\u3082\u81ea\u7531\u3002LiveCodeBench \u306e\u30b9\u30b3\u30a2\u306f\u4e0a\u4f4d\u30af\u30e9\u30b9\u3067\u30011M\u30b3\u30f3\u30c6\u30ad\u30b9\u30c8\u3092\u6d3b\u304b\u3057\u305f\u9577\u5927\u306a\u30b3\u30fc\u30c9\u30ec\u30d3\u30e5\u30fc\u3084\u5927\u898f\u6a21\u30ea\u30dd\u30b8\u30c8\u30ea\u3078\u306e\u5bfe\u5fdc\u304c\u8a55\u4fa1\u3055\u308c\u3066\u3044\u308b\u3002<\/p>\n<table>\n<thead>\n<tr>\n<th>\u30e2\u30c7\u30eb<\/th>\n<th>\u7dcf\u30d1\u30e9\u30e1\u30fc\u30bf<\/th>\n<th>\u30a2\u30af\u30c6\u30a3\u30d6<\/th>\n<th>\u30b3\u30f3\u30c6\u30ad\u30b9\u30c8<\/th>\n<th>SWE-bench<\/th>\n<th>\u30e9\u30a4\u30bb\u30f3\u30b9<\/th>\n<\/tr>\n<\/thead>\n<tbody>\n<tr>\n<td><a href=\"https:\/\/kimi.com\/code\" target=\"_blank\" rel=\"noopener\">Kimi K2.7 Code<\/a><\/td>\n<td>1T<\/td>\n<td>32B<\/td>\n<td>256K<\/td>\n<td>Pro 58.6 \/ Verified 80.2<\/td>\n<td>Modified MIT<\/td>\n<\/tr>\n<tr>\n<td><a href=\"https:\/\/qwenlm.github.io\/\" target=\"_blank\" rel=\"noopener\">Qwen3-Coder 480B<\/a><\/td>\n<td>480B<\/td>\n<td>35B<\/td>\n<td>128K+<\/td>\n<td>Verified 69.6%<\/td>\n<td>Apache-2.0<\/td>\n<\/tr>\n<tr>\n<td><a href=\"https:\/\/www.deepseek.com\/\" target=\"_blank\" rel=\"noopener\">DeepSeek V4-Flash<\/a><\/td>\n<td>284B<\/td>\n<td>13B<\/td>\n<td>1M<\/td>\n<td>\u301c70%\uff08LiveCode\u4e0a\u4f4d\uff09<\/td>\n<td>MIT<\/td>\n<\/tr>\n<tr>\n<td><a href=\"https:\/\/mistral.ai\" target=\"_blank\" rel=\"noopener\">Devstral 2<\/a><\/td>\n<td>24B<\/td>\n<td>24B\uff08dense\uff09<\/td>\n<td>32K<\/td>\n<td>\u2014\uff08\u30a8\u30fc\u30b8\u30a7\u30f3\u30c8\u7279\u5316\uff09<\/td>\n<td>Apache-2.0<\/td>\n<\/tr>\n<tr>\n<td><a href=\"https:\/\/qwenlm.github.io\/\" target=\"_blank\" rel=\"noopener\">Qwen3.6 27B<\/a><\/td>\n<td>27B<\/td>\n<td>27B\uff08dense\uff09<\/td>\n<td>128K<\/td>\n<td>77.2% SWE-bench<\/td>\n<td>Apache-2.0<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<p>\u51fa\u5178: <a href=\"https:\/\/kilo.ai\/open-source-models\" target=\"_blank\" rel=\"noopener\">Kilo Code\u300cBest Open-Source Coding Models 2026\u300d<\/a> \/ <a href=\"https:\/\/www.morphllm.com\/best-open-source-llm\" target=\"_blank\" rel=\"noopener\">MorphLLM \u30e9\u30f3\u30ad\u30f3\u30b0<\/a><\/p>\n<h2>Ollama v0.30.8\u2014\u2014GGUF \u62e1\u5f35\u3068 Apple Silicon MLX \u6539\u5584<\/h2>\n<p>\u30ed\u30fc\u30ab\u30eb LLM \u5b9f\u884c\u30c4\u30fc\u30eb <a href=\"https:\/\/ollama.com\/\" target=\"_blank\" rel=\"noopener\"><strong>Ollama<\/strong><\/a> \u306f2026\u5e746\u670812\u65e5\u306b v0.30.8 \u3092\u30ea\u30ea\u30fc\u30b9\u3057\u305f\u3002<a href=\"https:\/\/github.com\/ggml-org\/llama.cpp\" target=\"_blank\" rel=\"noopener\">llama.cpp<\/a> \u7d4c\u7531\u3067\u306e GGUF \u30cf\u30fc\u30c9\u30a6\u30a7\u30a2\u30b5\u30dd\u30fc\u30c8\u304c\u62e1\u5f35\u3055\u308c\u3001Apple Silicon \u5411\u3051 MLX \u30a8\u30f3\u30b8\u30f3\u3082\u6539\u5584\u3055\u308c\u3066\u3044\u308b\u3002<\/p>\n<ul>\n<li><strong>\u4eca\u6708\u30e9\u30a4\u30d6\u30e9\u30ea\u8ffd\u52a0\u306e\u4e3b\u306a\u30e2\u30c7\u30eb<\/strong>: MiniMax M3\uff081M\u30b3\u30f3\u30c6\u30ad\u30b9\u30c8\u30fb\u30d3\u30b8\u30e7\u30f3\uff09\u3001NVIDIA Nemotron 3 Ultra\uff086\u67084\u65e5\uff09\u3001DeepSeek V4 Pro\u3001Qwen3.6<\/li>\n<li><strong>\u30c0\u30a6\u30f3\u30ed\u30fc\u30c9\u6570\u30c8\u30c3\u30d7<\/strong>: \u5f15\u304d\u7d9a\u304d Llama 4 Scout \u30d5\u30a1\u30df\u30ea\u30fc\u3001Llama 3.x<\/li>\n<li><strong>\u30b3\u30f3\u30b7\u30e5\u30fc\u30de GPU \u3067\u306e\u30d9\u30b9\u30c8\u30e2\u30c7\u30eb<\/strong>: Qwen3.6 27B\uff08Q4\u91cf\u5b50\u5316\u306724GB\u3001SWE-bench 77.2%\uff09<\/li>\n<\/ul>\n<p>\u30b0\u30e9\u30d5\u30a3\u30ab\u30eb GUI \u7248\u3067\u306f <a href=\"https:\/\/lmstudio.ai\/\" target=\"_blank\" rel=\"noopener\"><strong>LM Studio<\/strong><\/a>\uff08Element Labs\uff09\u304c Mac\/Windows \u5411\u3051\u306b\u5f15\u304d\u7d9a\u304d\u4eba\u6c17\u3067\u3001MCP \u30b5\u30fc\u30d0\u30fc\u3068\u306e\u7d71\u5408\u304c\u6a19\u6e96\u5316\u3057\u300c\u30ed\u30fc\u30ab\u30eb AI \u30a8\u30fc\u30b8\u30a7\u30f3\u30c8\u30d7\u30e9\u30c3\u30c8\u30d5\u30a9\u30fc\u30e0\u300d\u3068\u3057\u3066\u4f4d\u7f6e\u3065\u3051\u304c\u660e\u78ba\u306b\u306a\u3063\u3066\u3044\u308b\u3002Linux \u3067\u306f\u30b3\u30de\u30f3\u30c9\u30e9\u30a4\u30f3\u904b\u7528\u304c\u4e3b\u6d41\u306e\u4e2d\u3001Ollama \u516c\u5f0f GUI\uff08Mac\/Windows \u5411\u3051\uff09\u3082 v0.10 \u4ee5\u964d\u642d\u8f09\u3055\u308c\u3066\u3044\u308b\u3002<\/p>\n<p>\u51fa\u5178: <a href=\"https:\/\/www.promptquorum.com\/local-llms\/top-open-source-models-ollama\" target=\"_blank\" rel=\"noopener\">PromptQuorum\u300cOllama Juin 2026 v0.30.8\u300d<\/a> \/ <a href=\"https:\/\/fungies.io\/local-llm-inference-tools-guide-2026-3\" target=\"_blank\" rel=\"noopener\">Fungies.io \u30ed\u30fc\u30ab\u30eb LLM \u63a8\u8ad6\u30c4\u30fc\u30eb\u30ac\u30a4\u30c9 2026<\/a><\/p>\n<hr \/>\n<h2>\u4e3b\u306a\u60c5\u5831\u6e90<\/h2>\n<ul>\n<li><a href=\"https:\/\/x.com\/Kimi_Moonshot\/status\/2065377579130142937\" target=\"_blank\" rel=\"noopener\">Moonshot AI \u516c\u5f0f X \u2014 Kimi K2.7 Code \u30ea\u30ea\u30fc\u30b9\u544a\u77e5<\/a><\/li>\n<li><a href=\"https:\/\/www.vm0.ai\/en\/models\/kimi-k2-7-code\" target=\"_blank\" rel=\"noopener\">VM0 \u2014 Kimi K2.7 Code \u30e2\u30c7\u30eb\u8a73\u7d30<\/a><\/li>\n<li><a href=\"https:\/\/www.verdent.ai\/guides\/kimi-k2-6-developers\" target=\"_blank\" rel=\"noopener\">Verdent Guides \u2014 Kimi K2.6\/K2.7 \u958b\u767a\u8005\u5411\u3051\u89e3\u8aac<\/a><\/li>\n<li><a href=\"https:\/\/mistral.ai\/news\/ocr-4\" target=\"_blank\" rel=\"noopener\">Mistral AI \u516c\u5f0f\u30d6\u30ed\u30b0 \u2014 Mistral OCR 4<\/a><\/li>\n<li><a href=\"https:\/\/kilo.ai\/open-source-models\" target=\"_blank\" rel=\"noopener\">Kilo Code \u2014 Best Open-Source Coding Models 2026<\/a><\/li>\n<li><a href=\"https:\/\/www.morphllm.com\/best-open-source-llm\" target=\"_blank\" rel=\"noopener\">MorphLLM \u2014 The Best Open Source LLMs (2026)<\/a><\/li>\n<li><a href=\"https:\/\/pinggy.io\/blog\/best_open_source_self_hosted_llms_for_coding\" target=\"_blank\" rel=\"noopener\">Pinggy \u2014 Best Open Source Self-Hosted LLMs for Coding 2026<\/a><\/li>\n<li><a href=\"https:\/\/www.promptquorum.com\/local-llms\/top-open-source-models-ollama\" target=\"_blank\" rel=\"noopener\">PromptQuorum \u2014 Ollama June 2026 v0.30.8<\/a><\/li>\n<li><a href=\"https:\/\/fungies.io\/local-llm-inference-tools-guide-2026-3\" target=\"_blank\" rel=\"noopener\">Fungies.io \u2014 Local LLM Inference Tools Guide 2026<\/a><\/li>\n<\/ul>\n<p><small>\u5404\u30b5\u30fc\u30d3\u30b9\u306f\u63d0\u4f9b\u5143\u306e\u898f\u7d04\u306b\u5f93\u3063\u3066\u3054\u5229\u7528\u304f\u3060\u3055\u3044\u3002\u60c5\u5831\u306f\u5404\u60c5\u5831\u6e90\u306e\u516c\u958b\u6642\u70b9\u306e\u3082\u306e\u3067\u3059\u3002\u53ce\u96c6\u30fb\u8981\u7d04\u306fAI\u3001\u516c\u958b\u5224\u65ad\u306f\u4eba\u3002<\/small><\/p>\n","protected":false},"excerpt":{"rendered":"<p>Moonshot AI\u306eKimi K2.7 Code\u304cSWE-bench Pro 58.6\u3067\u30aa\u30fc\u30d7\u30f3\u30a6\u30a7\u30a4\u30c8\u6700\u9ad8\u6c34\u6e96\u3092\u8a18\u9332\u3001Mistral\u304c\u30bb\u30eb\u30d5\u30db\u30b9\u30c8OCR AI\u300cOCR 4\u300d\u3092\u516c\u958b\u3001Alibaba\u306eQwen3-Coder\u3068DeepSeek V4-Flash\u3082\u5b58\u5728\u611f\u2014\u20142026\u5e746\u6708\u306e\u30aa\u30fc\u30d7\u30f3\u30bd\u30fc\u30b9LLM\u30b3\u30fc\u30c7\u30a3\u30f3\u30b0\u6700\u524d\u7dda\u3092\u307e\u3068\u3081\u305f\u3002<\/p>\n","protected":false},"author":1,"featured_media":58,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"page_builder":"","footnotes":""},"categories":[46],"tags":[],"class_list":["post-272","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-aillm"],"_links":{"self":[{"href":"https:\/\/www.wetch.co.jp\/ai\/wp-json\/wp\/v2\/posts\/272","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.wetch.co.jp\/ai\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.wetch.co.jp\/ai\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.wetch.co.jp\/ai\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.wetch.co.jp\/ai\/wp-json\/wp\/v2\/comments?post=272"}],"version-history":[{"count":1,"href":"https:\/\/www.wetch.co.jp\/ai\/wp-json\/wp\/v2\/posts\/272\/revisions"}],"predecessor-version":[{"id":392,"href":"https:\/\/www.wetch.co.jp\/ai\/wp-json\/wp\/v2\/posts\/272\/revisions\/392"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.wetch.co.jp\/ai\/wp-json\/wp\/v2\/media\/58"}],"wp:attachment":[{"href":"https:\/\/www.wetch.co.jp\/ai\/wp-json\/wp\/v2\/media?parent=272"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.wetch.co.jp\/ai\/wp-json\/wp\/v2\/categories?post=272"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.wetch.co.jp\/ai\/wp-json\/wp\/v2\/tags?post=272"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}