The Open Source Community's DeepSeek Moment
A few days ago, while researching text-to-video models, I noticed something that diverged significantly from my prior understanding: in the directions of text-to-image and text-to-video, Chinese companies' presence in the open-source community is far weaker than in language models.
I had assumed Chinese models were already dominant across the board. But looking closer, that's not the case. This prompted me to review how the open-source community has changed over the past three years.
Model Open Source Is Not Software Open Source
First, something many people may not have considered.
With open-source software, once the source code is released, there are no secrets left—you can rebuild an identical copy from scratch. Models don't work that way. In most cases, open-sourcing a model means releasing the weights and inference scripts. The core elements—training methods, training data, engineering details—usually remain private.
What can you do with the weights? Deploy inference, fine-tune. Want to reproduce from scratch? Nearly impossible. So "open source" for models was never the same thing as "open source" for software.
This led to a long-standing debate from the early days: if model open source isn't the same as software open source, then what's the point?
The biggest benefit of open-source software is community collaboration—developers worldwide fixing bugs and adding features together. But when a model is open-sourced, far too few people or institutions have the capacity to participate in its actual development. You need massive compute, data, training infrastructure. The vast majority of people can only run inference with the weights.
At the time, Robin Li (Baidu's CEO) said open-source large models were meaningless, and for a while I thought he had a point. If collaborative development—the primary driving force—doesn't apply here, then what's the point of open-sourcing?
Later events answered that question.
The Age of LLaMA
When ChatGPT emerged at the end of 2022, large models entered the public consciousness. Before that, the open-source world was rather dull. OpenAI's GPT-2 was fully open-sourced in 2019, but by GPT-3 it had become an application-based API access model. I remember wanting to use GPT-3 at a hackathon and discovering I had to email for API access. It was essentially closed-source.
In 2023, the open-source community was basically dominated by Meta's LLaMA. LLaMA 1 came in February 2023, LLaMA 2 in July. Every time LLaMA released a new version, a batch of domestic models would announce upgrades—this was the so-called "hundred-model war," with the pace set by LLaMA.
Chinese models open-sourced at this stage were, frankly, for marketing. The released versions were all relatively small. Zhipu's GLM-6B was the earliest representative—many people's first exposure to on-premise large model deployment started with it. I remember a friend choosing a model at the time and wondering why he picked a Chinese one; he said it was from Tsinghua and had a decent reputation. Baichuan open-sourced a 14B model. 01.AI, founded by Kai-Fu Lee, open-sourced Yi-34B in November 2023—that was one of the larger Chinese open-source models at the time. The Shanghai AI Laboratory was also steadily working on the InternLM (Shusheng) series.
Everyone followed the same strategy: open-source the small ones for promotion, keep the large ones closed for commercialization.
Qwen Enters the Field
In 2024, this equilibrium was broken by Qwen.
Starting mid-2024, Alibaba's Qwen began releasing intensively, from models with a few dozen billion parameters up to 72B, all performing well and all open-sourced. The default assumption had been that large models stay closed-source, but suddenly someone was releasing highly capable large models outright.
LLaMA had significant international influence, but its Chinese capabilities were always lacking; practical use required additional training. Qwen was essentially ready to use out of the box for Chinese scenarios, and quickly replaced LLaMA's position in the Chinese open-source community.
By the end of 2024, the Qwen series had become the de facto standard for Chinese open-source models. Closed-source models felt pressure from open-source competition for the first time.
DeepSeek Flipped the Table
Qwen played the existing rules to perfection. DeepSeek changed the rules entirely.
DeepSeek entered in 2024 with a simple playbook: open-source upon release, accompanied by extremely thorough technical reports. At the end of 2024, V3 was released—hundreds of billions of parameters, excellent performance, open-sourced immediately upon release. Few people at that point had seen a model of that scale released so openly.
But what really exploded was R1 in January 2025.
OpenAI had just launched the O1 reasoning model in September 2024, and DeepSeek's R1 came out right before the Spring Festival. Its reasoning capabilities were remarkably close to the top closed-source models of the time—not quite equal, but the gap was surprisingly small. And it was fully open-sourced on day one.
The established order had been "small models open, large models closed." DeepSeek open-sourced something better than many companies' best closed-source models, and that order collapsed overnight.
LLaMA 4 is another footnote. Meta trained a massive model for a long time, trying to reclaim its position in the open-source community, and launched it in April 2025—it flopped. The results fell far short of expectations, and benchmark cheating was exposed. Later, Yann LeCun himself admitted that "results were fudged", and Zuckerberg lost confidence in the entire GenAI team. LLaMA 4 was barely used, marking the end of the LLaMA series' standing in the open-source community.
Day-0 Open Source Became the Industry Norm
After DeepSeek, Chinese model companies shifted to day-0 open source one after another—their best models were open-sourced on the release day itself.
Kimi open-sourced K2 in July 2025, a trillion-parameter MoE model. MiniMax open-sourced M2.5. Zhipu continued iterating the GLM series. Wave after wave, the quality of open-source models kept rising.
Today, if you look at the international open-source community for language models, the leaderboards are almost entirely Chinese models. Qwen, DeepSeek, GLM, MiniMax, Kimi—the presence of overseas models has become very weak.
What DeepSeek did wasn't merely contribute a model. It changed how the entire industry plays its hand.
But the Winds Are Shifting Again
However, the fervor around this wave of day-0 open sourcing is cooling.
DeepSeek's last open-source release was V3.2 in December 2025—over four months ago. V4 has been rumored for a long time but hasn't materialized. During this gap, strategies have started to waver again.
Qwen 3.6 Plus was released at the end of March 2026 without open-sourcing, API-only. This was the first time a flagship Qwen model wasn't open-sourced. Zhipu's GLM-5.1 was also released closed-source first, though it just announced weight open-sourcing in the past couple of days. Many companies' latest multimodal models are no longer open-sourced either.
It seems we're back to the same old question: what's the point of open-sourcing? When competitive pressure eases, the answer may change again.
Text-to-Image and Text-to-Video Are Still Waiting
Back to the surprising discovery that started all this.
The text-to-image open-source community is still dominated by overseas models. The most widely used are Stability AI's Stable Diffusion series and Black Forest Labs' FLUX series. Chinese models have made some progress: Qwen released Qwen-Image, Tencent has Hunyuan Image 3.0, Zhipu has GLM-Image. But compared to the language model landscape, the difference is vast.
The same goes for text-to-video. Alibaba's last open-source text-to-video model was Wan 2.2 in July 2025—nearly nine months ago with nothing new since. The most buzzworthy open-source text-to-video model recently is LTX-2 from Israeli company Lightricks, which open-sourced its weights in January 2026.
This is a completely different world from language models. On the language model side, Chinese models have saturated the entire open-source community. On the text-to-image and text-to-video side, it still looks more like 2023: overseas models dominate, Chinese models appear sporadically. The entire open-source community is undergoing a "short-video-style" explosion, but this explosion is currently concentrated in language models and software tools.
What Are We Waiting For
Everyone is waiting for DeepSeek V4.
But it's not just a model we're waiting for. DeepSeek previously proved something: when a sufficiently strong model is fully open-sourced on its release day, it can shift the strategic direction of an entire industry. And from a business perspective, the endgame for model companies is becoming cloud companies—the business behind open-sourcing models is far larger than imagined. This happened in language models; it has not yet happened in text-to-image and text-to-video.
I sometimes half-jokingly think that maybe we just need DeepSeek to do it again. But then again, DeepSeek itself hasn't released a new model in over four months either.
Originally published at https://guanjiawei.ai/en/blog/open-source-deepseek-moment