Open-source AI models have closed the capability gap with proprietary systems like ChatGPT far more than the market revenue suggests. According to Mozilla's inaugural State of Open Source AI report, open-weight models now trail frontier closed models by just 3.3 percentage points on leading benchmarks, yet capture only 4% of global AI market revenue despite powering roughly one-third of all active AI applications.

Why Are Developers Using Open Models But Not Shipping Them?

The report surveyed more than 950 developers globally and uncovered a striking paradox: 79% of developers use open-weight models, but only 51% have deployed them in production, compared with 63% for closed models. This deployment gap reveals that the barrier to adoption is not raw intelligence or capability. Instead, it comes down to operational friction.

Raffi Krikorian, Chief Technology Officer of Mozilla, authored the report and identified what he calls the "agentic harness" problem as the core issue. This refers to the software layer between users and models that controls what an AI system can see, remember, and act on. According to the report, "changing the surrounding software can affect performance more than switching the model itself". Closed models come with managed application programming interfaces (APIs), which are standardized ways for software to communicate with the model. Open models require teams to build and maintain this infrastructure themselves.

What Specific Tasks Favor Open Models Over Closed Ones?

The 3.3-point capability gap has become small enough that organizations running straightforward inference workloads can often substitute open models without measurable quality loss. Closed models retain meaningful leads in three specific capability categories: reasoning, long-context retrieval, and complex agentic tasks. Outside those categories, open models are competitive for summarization, classification, and structured data extraction.

The economic case for open models has strengthened significantly. AI model costs have fallen up to 50 times over the past three years, narrowing the financial argument for paying the closed-model premium in cost-sensitive applications.

How to Evaluate Open Models for Your Organization

• Assess Your Engineering Capacity: Teams with strong engineering resources can increasingly use open models at substantially lower per-token cost, while teams without that infrastructure will continue to pay the closed-model premium for the managed API experience.
• Identify Your Use Case: If your workload involves straightforward inference tasks like summarization, classification, or structured extraction, open models may deliver equivalent quality at lower cost; if your application requires advanced reasoning or long-context retrieval, closed models may still offer meaningful advantages.
• Plan for Operational Overhead: Open models require investment in infrastructure setup, fine-tuning pipelines, security audits, and maintaining model versions, so budget for these operational costs when comparing total cost of ownership.

For business decision-makers, the evaluation question has fundamentally shifted. The question is no longer "are open models good enough?" but "does my team have the operational capacity to run them?".

Who Is Making Money on Open-Source AI?

Despite the revenue gap, a cluster of commercial open-source AI companies is growing rapidly. Databricks is running at a 5.4 billion dollar annual revenue run-rate. Mistral's annual recurring revenue reached approximately 400 million dollars, a 20-fold increase in 12 months. DeepSeek reported approximately 220 million dollars in annual recurring revenue and recently raised 7.4 billion dollars at a valuation above 50 billion dollars; its V4 model is processing 18.4 trillion tokens per month.

Regional adoption patterns show sharp divergence. China and East Asia report 89% open-source AI adoption, the highest of any region, with China treating open-source as an explicit national strategy.

"We have been here before. Mozilla exists because one company tried to own the front door to the web, and an open community rose up to make sure it never could. Twenty-five years later, someone is running the same play," said Raffi Krikorian.

Raffi Krikorian, Chief Technology Officer at Mozilla

The 4% revenue figure represents a market structural lag, not a ceiling. As tooling matures to close the operational gap, the revenue share for open models is expected to rise, making the current environment a transitional window for businesses that invest now in open-model infrastructure. Experts estimate this shift could occur within 12 to 18 months if the trajectory in Mozilla's data continues.

For organizations evaluating AI tools, the narrowing capability gap combined with dramatically lower costs means open models deserve serious consideration. The remaining competitive moat for closed models is the managed agentic harness, not raw intelligence. Teams that build the operational infrastructure now will have significantly lower inference costs when the supporting tooling catches up.