{"id":661524,"date":"2025-04-08T23:05:13","date_gmt":"2025-04-08T20:05:13","guid":{"rendered":"https:\/\/en.buradabiliyorum.com\/deep-cogito-emerges-from-stealth-with-hybrid-ai-reasoning-models\/"},"modified":"2025-04-08T23:05:13","modified_gmt":"2025-04-08T20:05:13","slug":"deep-cogito-emerges-from-stealth-with-hybrid-ai-reasoning-models","status":"publish","type":"post","link":"https:\/\/buradabiliyorum.com\/en\/deep-cogito-emerges-from-stealth-with-hybrid-ai-reasoning-models\/","title":{"rendered":"Deep Cogito emerges from stealth with hybrid AI &#8216;reasoning&#8217; models"},"content":{"rendered":"<div>\n<p id=\"speakable-summary\" class=\"wp-block-paragraph\">A new company, <a rel=\"nofollow\" target=\"_blank\" href=\"https:\/\/www.deepcogito.com\/\" target=\"_blank\" rel=\"noreferrer noopener nofollow\">Deep Cogito<\/a>, has emerged from stealth with a family of openly available AI models that can be switched between \u201creasoning\u201d and non-reasoning modes.<\/p>\n<p class=\"wp-block-paragraph\">Reasoning models like OpenAI\u2019s o1 have shown great promise in domains like math and physics, thanks to their ability to effectively fact-check themselves by working through complex problems step by step. This reasoning comes at a cost, however: higher computing and latency. That\u2019s why labs like Anthropic are pursuing \u201chybrid\u201d model architectures that combine reasoning components with standard, non-reasoning elements. Hybrid models can quickly answer simple questions while spending additional time considering more challenging queries.  <\/p>\n<p class=\"wp-block-paragraph\">All of Deep Cogito\u2019s models, called Cogito 1, are hybrid models. Cogito claims that they outperform the best open models of the same size, including models from Meta and Chinese AI startup DeepSeek.<\/p>\n<p class=\"wp-block-paragraph\">\u201cEach model can answer directly [\u2026] or self-reflect before answering (like reasoning models),\u201d the company <a rel=\"nofollow\" target=\"_blank\" href=\"https:\/\/www.deepcogito.com\/research\/cogito-v1-preview\" target=\"_blank\" rel=\"noreferrer noopener nofollow\">explained in a blog post<\/a>. \u201c[All] were developed by a small team in <a href=\"https:\/\/buradabiliyorum.com\/en\/category\/download-scripts-themes-apps\/\" data-internallinksmanager029f6b8e52c=\"9\" title=\"Download Scripts &amp; Themes &amp; Apps\" target=\"_blank\" rel=\"noopener\">app<\/a>roximately 75 days.\u201d<\/p>\n<p class=\"wp-block-paragraph\">The Cogito 1 models range from 3 billion parameters to 70 billion parameters, and Cogito says that models ranging up to 671 billion parameters will join them in the coming weeks and months. Parameters roughly correspond to a model\u2019s problem-solving skills, with more parameters <a href=\"https:\/\/buradabiliyorum.com\/en\/category\/general\/\" data-internallinksmanager029f6b8e52c=\"3\" title=\"General\" target=\"_blank\" rel=\"noopener\">general<\/a>ly being better.<\/p>\n<p class=\"wp-block-paragraph\">Cogito 1 wasn\u2019t developed from scratch, to be clear. Deep Cogito built on top of Meta\u2019s open Llama and Alibaba\u2019s Qwen models to create its own. The company says that it applied novel training approaches to boost the base models\u2019 performance and enable toggleable reasoning. <\/p>\n<p class=\"wp-block-paragraph\">According to the results of Cogito\u2019s internal benchmarking, the largest Cogito 1 model, Cogito 70B, with reasoning outperforms DeepSeek\u2019s R1 reasoning model on a few mathematics and language evaluations. Cogito 70B with reasoning disabled also eclipses Meta\u2019s recently released Llama 4 Scout model on LiveBench, a general-purpose AI test.<\/p>\n<p class=\"wp-block-paragraph\">Every Cogito 1 model is available for download or use via APIs on cloud providers Fireworks AI and Together AI.<\/p>\n<figure class=\"wp-block-image aligncenter size-large\"><img loading=\"lazy\" decoding=\"async\" width=\"1422\" height=\"800\" src=\"https:\/\/techcrunch.com\/wp-content\/uploads\/2025\/04\/Livebench.webp?w=680\" alt=\"Deep Cogito\" class=\"wp-image-2991208\" srcset=\"https:\/\/techcrunch.com\/wp-content\/uploads\/2025\/04\/Livebench.webp 1422w, https:\/\/techcrunch.com\/wp-content\/uploads\/2025\/04\/Livebench.webp?resize=150,84 150w, https:\/\/techcrunch.com\/wp-content\/uploads\/2025\/04\/Livebench.webp?resize=300,169 300w, https:\/\/techcrunch.com\/wp-content\/uploads\/2025\/04\/Livebench.webp?resize=768,432 768w, https:\/\/techcrunch.com\/wp-content\/uploads\/2025\/04\/Livebench.webp?resize=680,383 680w, https:\/\/techcrunch.com\/wp-content\/uploads\/2025\/04\/Livebench.webp?resize=1200,675 1200w, https:\/\/techcrunch.com\/wp-content\/uploads\/2025\/04\/Livebench.webp?resize=1280,720 1280w, https:\/\/techcrunch.com\/wp-content\/uploads\/2025\/04\/Livebench.webp?resize=430,242 430w, https:\/\/techcrunch.com\/wp-content\/uploads\/2025\/04\/Livebench.webp?resize=720,405 720w, https:\/\/techcrunch.com\/wp-content\/uploads\/2025\/04\/Livebench.webp?resize=900,506 900w, https:\/\/techcrunch.com\/wp-content\/uploads\/2025\/04\/Livebench.webp?resize=800,450 800w, https:\/\/techcrunch.com\/wp-content\/uploads\/2025\/04\/Livebench.webp?resize=668,375 668w, https:\/\/techcrunch.com\/wp-content\/uploads\/2025\/04\/Livebench.webp?resize=1097,617 1097w, https:\/\/techcrunch.com\/wp-content\/uploads\/2025\/04\/Livebench.webp?resize=708,398 708w\" sizes=\"auto, (max-width: 1422px) 100vw, 1422px\"\/><figcaption class=\"wp-element-caption\"><span class=\"wp-element-caption__text\">Cogito 1\u2019s performance compared to other popular openly available AI models<\/span><span class=\"wp-block-image__credits\"><strong>Image Credits:<\/strong>Deep Cogito<\/span><\/figcaption><\/figure>\n<p class=\"wp-block-paragraph\">\u201cCurrently, we\u2019re still in the early stages of [our] scaling curve, having used only a fraction of compute typically reserved for traditional large language model post\/continued training,\u201d wrote Cogito in its blog post. \u201cMoving forward, we\u2019re investigating complementary post-training approaches for self-improvement.\u201d<\/p>\n<p class=\"wp-block-paragraph\"><a rel=\"nofollow\" target=\"_blank\" href=\"https:\/\/www.bizprofile.net\/ca\/san-francisco\/deep-cogito-inc\" target=\"_blank\" rel=\"noreferrer noopener nofollow\">According to filings with California State<\/a>, San Francisco-based Deep Cogito was founded in June 2024. The company\u2019s <a rel=\"nofollow\" target=\"_blank\" href=\"https:\/\/www.linkedin.com\/company\/deep-cogito\/people\/\" target=\"_blank\" rel=\"noreferrer noopener nofollow\">LinkedIn page<\/a> lists two co-founders, Drishan Arora and Dhruv Malhotra. Malhotra was previously a product manager at Google AI lab DeepMind, where he worked on generative search <a href=\"https:\/\/buradabiliyorum.com\/en\/category\/technology\/\" data-internallinksmanager029f6b8e52c=\"4\" title=\"Technology\" target=\"_blank\" rel=\"noopener\">technology<\/a>. Arora was a senior software engineer at Google.<\/p>\n<p class=\"wp-block-paragraph\">Deep Cogito, whose backers include South Park Commons, <a rel=\"nofollow\" target=\"_blank\" rel=\"nofollow\" href=\"https:\/\/pitchbook.com\/profiles\/company\/697127-68#investors\">according to PitchBook<\/a>, ambitiously aims to build \u201cgeneral superintelligence.\u201d The company\u2019s founders understand the phrase to mean AI that can perform tasks better than most humans and \u201cuncover entirely new capabilities we have yet to imagine.\u201d<\/p>\n<\/div>\n<blockquote><p><strong><span style=\"color: #ff6600;\">If you liked the article, do not forget to share it with your friends. Follow us on\u00a0<span style=\"color: #ff0000;\"><a style=\"color: #ff0000;\" href=\"https:\/\/news.google.com\/publications\/CAAqBwgKMN63nwsw68G3Aw\" target=\"_blank\" rel=\"nofollow noopener noreferrer\">Google News<\/a><\/span>\u00a0too, click on the star and choose us from your favorites.<\/span><\/strong><\/p><\/blockquote>\n<blockquote>\n<p style=\"text-align: center;\"><strong>If you want to read more like this article, you can visit our <span style=\"color: #ff9900;\"><a style=\"color: #ff9900;\" href=\"https:\/\/en.buradabiliyorum.com\/category\/technology\/\" target=\"_blank\" >Technology<\/a><\/span> category.<\/strong><\/p>\n<\/blockquote>\n<p><span style=\"color: black;\"><a style=\"color: #ff9900;\" href=\"https:\/\/techcrunch.com\/2025\/04\/08\/deep-cogito-emerges-from-stealth-with-hybrid-ai-reasoning-models\/\" target=\"_blank\" >Source<\/a><\/span><\/p>\n","protected":false},"excerpt":{"rendered":"<p>A new company, Deep Cogito, has emerged from stealth with a family of openly available AI models that can be switched between \u201creasoning\u201d and non-reasoning modes. Reasoning models like OpenAI\u2019s o1 have shown great promise in domains like math and physics, thanks to their ability to effectively fact-check themselves by working through complex problems step&#8230;<\/p>\n","protected":false},"author":1,"featured_media":661525,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"fifu_image_url":"https:\/\/techcrunch.com\/wp-content\/uploads\/2025\/01\/GettyImages-1333209932.jpg?resize=1200,600","fifu_image_alt":"","footnotes":""},"categories":[18],"tags":[77337,155446,153095,70917],"class_list":["post-661524","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-technology","tag-ai","tag-deep-cogito","tag-reasoning","tag-startups"],"_links":{"self":[{"href":"https:\/\/buradabiliyorum.com\/en\/wp-json\/wp\/v2\/posts\/661524","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/buradabiliyorum.com\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/buradabiliyorum.com\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/buradabiliyorum.com\/en\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/buradabiliyorum.com\/en\/wp-json\/wp\/v2\/comments?post=661524"}],"version-history":[{"count":0,"href":"https:\/\/buradabiliyorum.com\/en\/wp-json\/wp\/v2\/posts\/661524\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/buradabiliyorum.com\/en\/wp-json\/wp\/v2\/media\/661525"}],"wp:attachment":[{"href":"https:\/\/buradabiliyorum.com\/en\/wp-json\/wp\/v2\/media?parent=661524"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/buradabiliyorum.com\/en\/wp-json\/wp\/v2\/categories?post=661524"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/buradabiliyorum.com\/en\/wp-json\/wp\/v2\/tags?post=661524"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}