The DeepSeek Moment for Open Source
A few days ago, while researching text-to-video models, I realized something that diverged significantly from my prior understanding: in 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 dominant across the board. Turns out, 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—you can build an identical copy from scratch. Models are different. In most cases, open-sourcing a model means releasing the weights and inference scripts, while the core elements—training methods, training data, engineering details—usually remain private.
What can you do with the weights? Deploy inference, fine-tune. Reproduce from scratch? Almost impossible. So "open source" for models and "open source" for software were never the same thing to begin with.
This led to a long-standing debate: since model open source differs from software open source, what's the point of open-sourcing models?
The greatest benefit of open-source software is community collaboration—developers worldwide fixing bugs and adding features together. But after a model is open-sourced, very few individuals or institutions have the capacity to participate in its actual development. You need massive compute, data, and training infrastructure. Most people can only run inference with the weights.
At the time, Robin Li (founder of Baidu) said open-source large models were meaningless, and for a while I thought he had a point. If community collaboration—the biggest driving force—doesn't apply, what's the point of open-sourcing?
Subsequent 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 GPT-3 shifted to an application-based API access model. I remember wanting to use GPT-3 at a hackathon and discovering I had to email to request API access. It was essentially closed-source.
In 2023, Meta's LLaMA basically dominated the open-source community. LLaMA 1 came in February 2023, LLaMA 2 in July. Every time LLaMA released a new version, a wave of domestic Chinese models would announce upgrades—this was the so-called "hundred-model war," with the tempo set by LLaMA.
Chinese models open-sourced at this stage were essentially for marketing purposes. The released versions were relatively small. Zhipu AI's GLM-6B was the earliest representative; for many people, it was their first exposure to on-premise large model deployment. 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, and Kai-Fu Lee's 01.AI open-sourced Yi-34B in November 2023, which was relatively large for a Chinese open-source model at the time. The Shanghai Artificial Intelligence Laboratory also continued its InternLM (Shusheng) series.
Everyone followed the same strategy: open-source the small ones for promotion, keep the large ones closed for commercialization.
Qwen Enters
In 2024, this equilibrium was broken by Qwen.
Starting in mid-2024, Alibaba's Qwen intensified its efforts, releasing models ranging from a few dozen billion to 72B parameters, all with strong performance and fully open-sourced. The previous assumption was that large models wouldn't be open-sourced; suddenly, someone was releasing highly capable large models outright.
Although LLaMA had significant international influence, its Chinese capabilities were always lacking, requiring secondary training for practical use. Qwen worked almost 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. For the first time, closed-source models felt pressure from open-source alternatives.
DeepSeek Flipped the Table
Qwen played the existing game perfectly. DeepSeek changed the rules entirely.
DeepSeek entered in 2024 with a simple strategy: open-source immediately, accompanied by exceptionally thorough technical reports. At the end of 2024, V3 was released—hundreds of billions of parameters, extremely capable, open-sourced on day one. Few people had seen a model of that scale released openly at that point.
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 before the Spring Festival. Its reasoning capabilities were extremely close to the top closed-source models at the time—not quite on par, but the gap was surprisingly small. And it was fully open-sourced on day one.
The old order was "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 spent a long time training a massive model to reclaim its position in the open-source community, releasing it in April 2025, only to crash and burn. Its performance fell far short of expectations, and it was embroiled in a benchmark-cheating scandal. Later, Yann LeCun himself admitted "results were fudged," and Zuckerberg lost confidence in the entire GenAI team. LLaMA 4 was basically unused, marking the end of the LLaMA series' standing in the open-source community.
Day-0 Open Source Became the Norm
After DeepSeek, Chinese model companies shifted to day-0 open source one after another, releasing their best models on the same day they were announced.
In July 2025, Kimi open-sourced K2, a trillion-parameter MoE model. MiniMax open-sourced M2.5. Zhipu AI 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—overseas models have become barely noticeable.
What DeepSeek did wasn't just contribute a model. It changed how the entire industry plays its cards.
But the Winds Are Shifting Again
However, the day-0 open-source fervor is cooling down.
DeepSeek's last open-source release was V3.2 in December 2025—more than four months ago. V4 has been rumored for a long time but hasn't materialized. During this lull, strategies have started to loosen.
Qwen 3.6 Plus was released at the end of March 2026 without being open-sourced—API-only. This was the first flagship Qwen model not to be open-sourced. Zhipu AI'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 being open-sourced either.
It seems we're back to that old question: what's the point of open-sourcing? When competitive pressure decreases, the answer may change again.
Text-to-Image and Text-to-Video Are Still Waiting
Returning to the finding that surprised me at the beginning.
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, and Zhipu AI has GLM-Image. But compared to the language model situation, the gap 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 recently buzzworthy open-source text-to-video model is LTX-2, from Israeli company Lightricks, which open-sourced its weights in January 2026.
It's a completely different world from language models. On the language model side, Chinese models have saturated the entire open-source community, while text-to-image and text-to-video still look more like 2023: overseas models dominate, with Chinese models appearing only sporadically. The entire open-source community is experiencing a "short-video-style" explosion, but this explosion has mainly occurred in language models and software tools so far.
What Are We Waiting For
Everyone is waiting for DeepSeek V4.
But we're waiting for more than just a model. DeepSeek previously proved something: a sufficiently powerful model fully open-sourced on release day can shift an entire industry's strategic direction. 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 with language models, but not yet with text-to-image and text-to-video.
I sometimes half-jokingly think that maybe DeepSeek just needs 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