{"id":54261,"date":"2025-11-29T08:19:51","date_gmt":"2025-11-29T08:19:51","guid":{"rendered":"https:\/\/youzum.net\/what-to-be-thankful-for-in-ai-in-2025\/"},"modified":"2025-11-29T08:19:51","modified_gmt":"2025-11-29T08:19:51","slug":"what-to-be-thankful-for-in-ai-in-2025","status":"publish","type":"post","link":"https:\/\/youzum.net\/fr\/what-to-be-thankful-for-in-ai-in-2025\/","title":{"rendered":"What to be thankful for in AI in 2025"},"content":{"rendered":"

Hello, dear readers. Happy belated Thanksgiving and Black Friday!<\/p>\n

This year has felt like living inside a permanent DevDay. Every week, some lab drops a new model, a new agent framework, or a new \u201cthis changes everything\u201d demo. It\u2019s overwhelming. But it\u2019s also the first year I\u2019ve felt like AI is finally diversifying \u2014 not just one or two frontier models in the cloud, but a whole ecosystem: open and closed, giant and tiny, Western and Chinese, cloud and local.<\/p>\n

So for this Thanksgiving edition, here\u2019s what I\u2019m genuinely thankful for in AI in 2025 \u2014 the releases that feel like they\u2019ll matter in 12\u201324 months, not just during this week\u2019s hype cycle.<\/p>\n

1. OpenAI kept shipping strong: GPT-5, GPT-5.1, Atlas, Sora 2 and open weights<\/b><\/h3>\n

As the company that undeniably birthed the “generative AI” era with its viral hit product ChatGPT in late 2022, OpenAI arguably had among the hardest tasks of any AI company in 2025: continue its growth trajectory even as well-funded competitors like Google with its Gemini models and other startups like Anthropic fielded their own highly competitive offerings. <\/p>\n

Thankfully, OpenAI rose to the challenge and then some. Its headline act was GPT-5, unveiled in August as the next frontier reasoning model, followed in November by GPT-5.1 <\/a>with new Instant and Thinking variants that dynamically adjust how much \u201cthinking time\u201d they spend per task. <\/p>\n

In practice, GPT-5\u2019s launch was bumpy \u2014 VentureBeat documented early math and coding failures and a cooler-than-expected community reaction in \u201cOpenAI\u2019s GPT-5 rollout is not going smoothly<\/a>,” but it quickly course corrected based on user feedback and, as a daily user of this model, I’m personally pleased with it and impressed with it. <\/p>\n

At the same time, enterprises actually using the models are reporting solid gains. ZenDesk Global<\/a>, for example, says GPT-5-powered agents now resolve more than half of customer tickets<\/a>, with some customers seeing 80\u201390% resolution rates. That\u2019s the quiet story: these models may not always impress the chattering classes on X, but they\u2019re starting to move real KPIs.<\/p>\n

On the tooling side, OpenAI finally gave developers a serious AI engineer with GPT-5.1-Codex-Max, a new coding model that can run long, agentic workflows and is already the default in OpenAI\u2019s Codex environment. VentureBeat covered it in detail in \u201cOpenAI debuts GPT-5.1-Codex-Max coding model and it already completed a 24-hour task internally<\/a>.\u201d <\/p>\n

Then there\u2019s ChatGPT Atlas, a full browser with ChatGPT baked into the chrome itself<\/a> \u2014 sidebar summaries, on-page analysis, and search tightly integrated into regular browsing. It\u2019s the clearest sign yet that \u201cassistant\u201d and \u201cbrowser\u201d are on a collision course.<\/p>\n

On the media side, Sora 2 turned the original Sora video demo into a full video-and-audio model with better physics, synchronized sound and dialogue, and more control over style and shot structure, plus a dedicated Sora app<\/a> with a full fledged social networking component, allowing any user to create their own TV network in their pocket<\/a>. <\/p>\n

Finally \u2014 and maybe most symbolically \u2014 OpenAI released gpt-oss-120B and gpt-oss-20B<\/a>, open-weight MoE reasoning models under an Apache 2.0\u2013style license. Whatever you think of their quality (and early open-source users have been loud about their complaints), this is the first time since GPT-2 that OpenAI has put serious weights into the public commons.<\/p>\n

2. China\u2019s open-source wave goes mainstream<\/b><\/h3>\n

If 2023\u201324 was about Llama and Mistral, 2025 belongs to China\u2019s open-weight ecosystem.<\/p>\n

A study from MIT and Hugging Face found that China now slightly leads the U.S. in global open-model downloads<\/a>, largely thanks to DeepSeek and Alibaba\u2019s Qwen family. <\/p>\n

Highlights:<\/p>\n