Open vs. Proprietary AI: What the Next 12 Months of Model Competition Could Look Like
The artificial intelligence (AI) world has seen a shift in momentum. Once-dominant proprietary models from the likes of OpenAI and Anthropic now face real competition from open-source challengers like DeepSeek, Llama and Mixtral. These open-weight models are no longer just alternatives for hobbyists—they’re becoming viable tools for real-world deployment across startups, research labs and even enterprise settings.
Research from McKinsey, the Mozilla Foundation and the Patrick J. McGovern Foundation found that placing a high priority on AI and using open-source tools go hand in hand: Organizations that view AI as important to their competitive advantage are over 40% more likely to report using open-source AI models and tools.
This changing landscape has triggered a new kind of “arms race”—one not solely about accuracy or capability but about control, cost and customization. And it’s not just technical teams paying attention. Business leaders and innovation officers are increasingly weighing whether to stick with polished, closed ecosystems or bet on open-source platforms that promise flexibility and transparency.
To understand where things are headed, we turned to members of the Senior Executive AI Think Tank—a group of leading AI strategists, engineers and product executives. Their collective view? The next 12 months won’t bring a clear winner. But they will bring clarity about where each type of model fits—and what that means for innovation, regulation and competitive edge.
A Race, Not a Knockout Fight
Divya Parekh, founder of The DP Group, doesn’t see this as a zero-sum battle. Instead, she describes the unfolding dynamic as a “tug-of-war” between community-powered open models and heavily resourced proprietary players.
“It’s not a knockout fight. It’s a dynamic dance—open and closed systems are shaping each other in real time.”
Open models, she notes, are advancing quickly thanks to rapid iteration, flexibility and lower costs—qualities that are especially appealing to agile organizations trying to innovate without overspending. But at the same time, large proprietary platforms are doubling down on the very features that give them an edge in highly regulated industries: security, compliance and dependable performance at scale.
“It won’t be a knockout fight,” says Parekh. “It’s more of a dynamic dance—both sides are pushing each other forward, shaping the AI landscape in real time.”
Strategic Platforms vs. Turnkey Ecosystems
Roman Vinogradov, VP of Product at Improvado, sees the real split emerging not between open and closed but between modular versus turnkey systems.
According to Vinogradov, open models like Mixtral and Llama aren’t just budget-friendly—they’re programmable infrastructure. That makes them a smart bet for dev-first teams that prioritize control, transparency and the ability to deploy AI where and how they want. Meanwhile, proprietary models are evolving into full-stack solutions offering polished user experience (UX), enterprise support and built-in compliance.
“The choice won’t be about performance alone,” Vinogradov explains. “It’s about who owns the logic—and how deeply AI gets embedded into a company’s DNA.”
Enterprise Adoption Will Stay Split
Mo Ezderman, Director of AI at MindGrub Technologies, believes the competition between open and proprietary models will become more segmented, not less. Open-source platforms will continue gaining ground in research and development (R&D), startups and international markets where cost and access are paramount. But large enterprises—especially those in finance, healthcare or government—will likely stick with vetted, closed systems for now.
“Open models will drive a wave of innovation fueled by budget constraints,” Ezderman says. “But proprietary models will retain a foothold in enterprise environments where support, compliance and risk mitigation matter most.”
The result, he predicts, will be coexistence: a maturing ecosystem where both model types raise the bar and push each other to evolve.
Open-Source Challenges—And Advantages
Gordon Pelosse, EVP of Partnerships and Enterprise Strategy at AI CERTs, is optimistic about the momentum behind open-source AI but notes that major hurdles remain. Open models excel in transparency and adaptability, but they still lag behind in standardized safety testing and performance benchmarks.
“Capabilities are commoditizing fast. That’s forcing every vendor to innovate and validate their value.”
That said, Pelosse anticipates rapid progress. He expects knowledge-sharing and public-private partnerships to help close the performance gap, even as proprietary vendors are forced to justify their pricing through real differentiators—like specialized vertical applications or superior UX.
“Capabilities will continue to commoditize,” he says. “That will force everyone to innovate faster.”
Use Case Will Dictate Choice
David Obasiolu, Co-Founder of Vliso AI, sees the battleground being defined less by ideology and more by industry use case.
He believes open-source models will thrive in sectors that prioritize customization, data privacy or edge deployment—areas where flexibility and control are non-negotiable. Meanwhile, companies focused on speed-to-value, scalability or simplified integration may continue leaning on proprietary offerings.
Competitive Gaps Are Narrowing
Jim Liddle, Chief Innovation Officer of Data Intelligence and AI at Nasuni, has been tracking the progress of open models closely. He sees the technical gap closing—fast.
“New open models aren’t just catching up,” Liddle notes. “They’re competing head-to-head with commercial foundational models—even on advanced features like multimodal capability and long context windows.”
As one example, Harvard Medical School shared research that found “open-source AI model performed on par with leading proprietary AI tool in solving tough medical cases that require complex clinical reasoning,” concluding that “findings suggest that open-source AI tools are becoming increasingly competitive and could offer a valuable alternative to proprietary models.”
“Open models are now competing on advanced features, not just cost. The performance gap is shrinking.”
This shift is forcing commercial vendors to respond. Liddle expects proprietary models to fight back with even larger models, enhanced reasoning tools (think GPT-5-level advances) or more intuitive AI experiences. In short, the next wave of breakthroughs is likely already underway.
What AI Model Competition Means for Business Leaders
The consensus from the AI Think Tank is clear: We’re not heading toward consolidation—we’re heading toward a richer, more diversified AI landscape.
For organizations investing in AI over the next 12 months, this means one thing: choice. The competition between open-source and proprietary models will continue to accelerate, driving better performance, lower costs and more strategic flexibility.
The key for decision-makers is understanding not just what’s available but what aligns best with your goals—whether that’s compliance, speed, cost, control or innovation. Because in the end, it’s not about which model is better; it’s about which one is better for you.
Takeaways for Business Leaders
- Expect coexistence, not consolidation. Open-source and proprietary models are evolving in parallel, not in opposition. Choosing one doesn’t mean rejecting the other.
- Let your use case guide your choice. Open models offer flexibility and customization; proprietary ones offer compliance and support. Your needs should determine your direction.
- Don’t overlook infrastructure strategy. Open models require more internal investment and expertise. Proprietary systems are faster to deploy but may offer less control.
- Prepare for faster innovation cycles. As competition heats up, both camps will push out updates, features and ecosystem improvements more rapidly. Stay nimble.
- Performance gaps are narrowing. Technical trade-offs between open and closed models are shrinking—especially on capabilities like long-context windows, multimodality and reasoning.
- Control is the next battleground. The conversation is shifting from “Which model performs better?” to “Who owns the logic, data and outcomes?”
