The Open Source AI Revolution: Driving Innovation, Efficiency, and Accessibility at Web Summit Lisbon 2026
<video style="width: 100%;height: auto" controls><source src="https://d2qu7nggdaeihr.cloudfront.net/output/67bf8048-752a-43d1-84fc-7a03cfdd45ad-1794308032.mp4" type="video/mp4"></video>
<aside>(This article was generated with AI and it&#8217;s based on a AI-generated transcription of a real talk on stage. While we strive for accuracy, we encourage readers to verify important information.)</aside>
<img decoding="async" src="https://web-summit-engine.imgix.net/fci4271v2dncgu040odtn2gc0hv6?auto=format&amp;fit=fill&amp;fill=solid&amp;fill-color=white&amp;w=600&amp;h=600&amp;fm=jpg&amp;dl=Reya El-Salahi.jpg" alt="Reya El-Salahi" align="left" hspace="10" width="200" style="float:left;margin-right: 15px">
At Web Summit Lisbon 2026, Presenter Reya El-Salahi moderated a panel on open source AI with Mr. Dahai Li (ModelBest), Mr. Alan Shen (MiniMax), and Mr. Rohit Sharma (True Ventures). Mr. Sharma noted AI&#8217;s evolution from closed to open ecosystems, mirroring past tech cycles. He anticipates open source and open weight AI models will become prevalent across the stack, creating significant investment opportunities.

Mr. Dahai Li championed smaller, efficient &#8220;edge models&#8221; like MiniCPM, arguing &#8220;bigger is not always better.&#8221; Businesses prioritize quality, speed, cost, and data control. His 7B model demonstrated superior real-world efficiency on limited hardware, lowering deployment barriers. Mr. Alan Shen affirmed openness as a core design, fostering an invaluable feedback loop from the global developer community, driving unforeseen innovations and strengthening MiniMax&#8217;s research.

For businesses, adopting open source models often requires sophisticated in-house tech teams, especially for on-premises deployment to address data privacy. While initial cost reductions are appealing, these can be balanced by additional security or engineering expenses. In-house AI is financially sensible for consistent, high usage or specific tasks, and critical when regulations prohibit external data transfer, suggesting a hybrid cloud-local approach.

The panelists envisioned a future with stratified AI models. Mr. Sharma predicted a mix of frontier models for complex tasks, customizable open-source models, and tiny models on personal devices. Mr. Shen emphasized matching the &#8220;right task for the right model&#8221; for optimal adoption. Addressing competition and safety, Mr. Shen stated that openness fuels innovation, citing MiniMax&#8217;s open video generation model as a foundation for numerous startups. Mr. Li argued open models are safer due to global independent testing and rapid vulnerability identification, akin to car safety features.

Looking ahead, Mr. Shen suggested that for open models to dominate corporate choices, better orchestration layers are needed to efficiently route routine tasks to open solutions, potentially saving millions. Mr. Sharma, while optimistic, expressed concern that immense compute capacity controlled by hyperscalers could limit the broader ecosystem&#8217;s ability to run open-source workloads at scale. Mr. Li concluded by noting AI&#8217;s accelerating evolution, foreseeing a future where powerful, universally accessible AI models integrate into every aspect of daily life within two to three years.

The open source advantage

The AI paradox: Smarter models, disappointing results

When AI builds itself