{"id":38821,"date":"2025-07-03T11:53:28","date_gmt":"2025-07-03T03:53:28","guid":{"rendered":"https:\/\/www.1ai.net\/?p=38821"},"modified":"2025-07-03T11:53:28","modified_gmt":"2025-07-03T03:53:28","slug":"%e6%99%ba%e8%b0%b1%e5%bc%80%e6%ba%90%e6%96%b0%e4%b8%80%e4%bb%a3%e9%80%9a%e7%94%a8%e8%a7%86%e8%a7%89%e8%af%ad%e8%a8%80%e6%a8%a1%e5%9e%8b","status":"publish","type":"post","link":"https:\/\/www.1ai.net\/en\/38821.html","title":{"rendered":"Smart Spectrum Open Source Next Generation Universal Visual Language Model"},"content":{"rendered":"

Yesterday,Zhipu<\/a>Officially launched andOpen Source<\/a>A new generation of general-purposevisual language model<\/a> GLM-4.1V-Thinking, which is claimed to be \"the key leap from perception to cognition achieved by the GLM family of visual models.<\/p>\n

\"Wisdom<\/p>\n

Specifically, GLM-4.1V-Thinking is a general-purpose inference-based grand model that supports multimodal inputs such as images, videos, documents, etc., and is designed for complex cognitive tasks. Based on the GLM-4V architecture, it introduces \"CoT Reasoning\" and adopts \"Reinforcement Learning Strategies for Lesson Sampling (RLCS)\", which systematically improves the model's cross-modal causal reasoning ability and stability.<\/p>\n

The lightweight version of the GLM-4.1V-9B-Thinking model is controlled at the 10B level, and it has achieved the best results of 10B in 23 out of 28 authoritative evaluations by MMStar, MMMU-Pro, and so on, of which 18 of them are equal to or better than the Qwen-2.5-VL, which has a parameter count as high as 72B.<\/p>\n

Officially, the GLM-4.1V-9B-Thinking demonstrates a high degree of versatility and robustness by excelling in five major areas: graphic comprehension, mathematical and scientific reasoning, video comprehension, GUI and web-intelligence body tasks, and visual anchoring and entity localization.<\/p>\n

Currently, GLM-4.1V-9B-Thinking has been open-sourced in GitHub, HuggingFace, and Magic Hitch communities, and published technical papers, API interface documents, and this time, two models, GLM-4.1V-9B-Base base model and GLM-4.1V-9B-Thinking are online.<\/p>\n

Link to paper:<\/strong>https:\/\/arxiv.org\/abs\/2507.01006<\/p>\n

GitHub:<\/strong>https:\/\/github.com\/THUDM\/GLM-4.1V-Thinking<\/p>\n

HuggingFace:<\/strong>https:\/\/huggingface.co\/collections\/THUDM\/glm-41v-thinking-6862bbfc44593a8601c2578d<\/p>\n

Magic Match Community:<\/strong>https:\/\/modelscope.cn\/collections\/GLM-41V-35d24b6def9f49<\/p>\n

API interface documentation:<\/strong>https:\/\/www.bigmodel.cn\/dev\/api\/visual-reasoning-model\/glm-4.1v-thinking<\/p>\n

Smart Spectrum has also launched a new eco-platform\"Agent Application Space<\/strong>.<\/p>\n

It is reported that \"Agent Application Space\" is an AI Agent capability aggregation platform for enterprise customers and developers, which aggregates rich Agent applications and Model Plug-ins (MCPs), provides out-of-the-box, flexible and orchestrated component services and Agents applications, and helps enterprises eliminate the need to build their own large model teams.<\/p>\n

In addition, on July 2, at the Smart Spectrum Open Platform Industry Ecological Conference, it was also announced that Pudong Venture Capital Group and Zhangjiang Group had made a total of 1 billion yuan of strategic investment in Smart Spectrum, and the first delivery was completed recently.<\/p>","protected":false},"excerpt":{"rendered":"

Yesterday, a new generation of GLM-4.1V-Thinking, known as the GLM series of visual models, was officially launched and opened. Specifically, GLM-4.1V-Thinking is a large generic reasoning model that supports multi-modular input of images, videos, documents, etc., designed for complex cognitive tasks. Based on the GLM-4V framework, it introduced the \"CoT Reassoning Mechanism\" and adopted the \"RLCS\" to systematically enhance the ability and stability of cross-modular causal reasoning models. Lightweight GLM-4.1V-9B-Thinking model parameters are controlled at 10B level<\/p>","protected":false},"author":1,"featured_media":0,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[146],"tags":[219,2680,4981],"collection":[],"class_list":["post-38821","post","type-post","status-publish","format-standard","hentry","category-news","tag-219","tag-2680","tag-4981"],"acf":[],"_links":{"self":[{"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/posts\/38821","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/comments?post=38821"}],"version-history":[{"count":0,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/posts\/38821\/revisions"}],"wp:attachment":[{"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/media?parent=38821"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/categories?post=38821"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/tags?post=38821"},{"taxonomy":"collection","embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/collection?post=38821"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}