For a few years the argument looked settled. The most capable models would be built behind closed doors by a handful of well-funded labs and rented out through an interface that revealed nothing of the machinery inside. Open alternatives — models whose weights anyone could download, inspect and run — were filed under worthy hobby: perpetually a generation behind, fine for tinkering, unserious for real work. That consensus has quietly collapsed. By mid-2026 the gap between the best open-weight models and the proprietary frontier has narrowed from a chasm to a margin, and the industry is rearranging itself around the new geometry.
The closed giants still hold the summit. For the hardest reasoning, the longest context and the most dependable output, the frontier labs remain ahead and mean to stay there. But the distance has shrunk to the point where, for the overwhelming majority of real tasks, an open model that costs a fraction as much to run is simply good enough — and "good enough, far cheaper, entirely under your own control" is one of the most persuasive sentences in enterprise software.
The case on each side
The argument for the closed approach is coherent, not merely self-serving. Frontier capability is genuinely expensive to build, and someone has to fund the compute and the talent; a subscription pays for the next leap. Concentration also makes safety tractable, the labs argue: a model reached only through an interface can be monitored, rate-limited and switched off if it misbehaves, whereas weights loose in the world can never be recalled. Keep the most powerful systems behind a controllable boundary and you keep a hand on the brake.
The open camp answers from a different set of values. A model you can download and run is one no vendor can deprecate, price-gouge or quietly alter beneath you. You can audit it, tune it on your own data, run it inside your own walls and know that nothing leaves the building. And the safety argument cuts both ways: open weights let thousands of independent researchers probe a model for flaws instead of trusting one company's private assurances. To this camp control is not something you cede to a vendor. It is something you refuse to hand over.
"The question stopped being 'which model is best.' It became 'which model do I actually control' — and for a great many companies that changes the answer entirely."
A machine-learning lead at a European bank — interviewed for this articleEnterprises simply run both
What strikes you, talking to the people who actually deploy these systems, is how few treat it as a binary. The sophisticated answer in 2026 is not open or closed but open and closed, routed by task. The pattern repeats across industry after industry: a closed frontier model for the genuinely hard, low-volume problems where capability earns its price and its dependency, and an open model running in-house for the high-volume, privacy-sensitive, cost-conscious work that makes up the bulk of the load.
The hedging is partly negotiation. An enterprise that can credibly run open models in production holds real leverage over its closed-model vendor; the threat of migration has stopped being theoretical. It is partly resilience — no single supplier can hold a critical workflow hostage. And it is partly arithmetic: for a task you run a million times a day, the gap between a frontier model's per-token price and an open model's near-zero marginal cost is the gap between a viable product and a money pit. The result is an architecture designed to refuse lock-in — and the same impulse toward independence runs well beyond software, as our investigation into Europe's decade-long sovereignty bet lays out.
Why Europe leans open
Nowhere is the open-model case more politically charged than in Europe, where it has fused with a broader unease about technological dependence. To policymakers in Brussels, a future in which every important AI system is rented from a handful of non-European firms looks uncomfortably like the chip dependency they are spending hundreds of billions to escape. Open weights offer a way out: a capability that can be hosted on the continent's own soil, audited by its own institutions and adapted to its own languages and rules, with no permanent tether to a foreign vendor's pricing or policy shifts.
That has turned open models into something close to industrial strategy. European research groups and startups have leaned into open releases not only on principle but as a competitive wedge — a way to matter in a field whose frontier they cannot yet out-spend. It is the same logic driving the continent's semiconductor push: you may not own the absolute leading edge, but a credible, controllable, home-grown capability is its own form of power, and one nobody abroad can revoke.
None of this means the closed labs are losing. They still define what is possible, still capture the most lucrative high-end work, still sit a generation ahead at the very top. But the war the headline promises is real, and it is being fought less over who builds the single best model than over the shape of the market itself — whether intelligence becomes a utility rented from a few towers or a commodity anyone can run. In mid-2026 both futures are live, and the players worth watching are quietly building for both.