Bolt.new, the AI-powered web development platform built by StackBlitz, has launched Forge, a research preview designed to make AI-assisted coding accessible to more people by removing the cost barrier. The new feature offers individual Pro subscribers 50 times more usage on open-source AI models at no additional cost, funded through an explicit data-sharing arrangement with Arcee AI, a leading open-weight model developer.

The economics of AI coding tools have traditionally favored those who can afford premium pricing. Bolt.new's founder and CEO Eric Simons explained the motivation behind Forge: "Price has been deciding who gets to build with AI. Plenty of people arrive at Bolt.new with an idea and the skill to see it through, and the cost of entry stops them there. Forge opens AI building to far more people".

Eric Simons

How Does Bolt Forge's Data-Sharing Model Work?

Forge operates on a transparent trade: developers who opt in share anonymized versions of their coding sessions, which Arcee AI uses to train a trillion-parameter-class open-weight model. The arrangement includes several safeguards to protect developer privacy and maintain control over the process.

• Consent Mechanism: Developers must actively opt in each time they switch to Forge mode, rather than agreeing once in buried terms of service.
• Data Anonymization: Bolt.new strips sensitive and personal information from sessions before any data leaves its infrastructure and validates the pipeline against test data.
• Legal Framework: A signed data processing agreement governs every transfer of information between Bolt.new and Arcee AI.
• Exit Option: Developers can leave the program at any time, and Teams and Enterprise workspaces are entirely excluded from data collection.

The circular economy closes when the trained models are released publicly. Developers whose work helped train the models can download, inspect, and run them like anyone else, creating what Arcee AI's co-founder and CEO Mark McQuade calls a path to broader ownership. "When a model's weights are open, capability stops being something you rent," McQuade stated. "Models trained on Forge are released to the public, so anyone can download and run them, including the builders whose work made them better".

Which AI Models Power Forge?

Forge runs on multiple open-source and semi-open models, giving developers flexibility in their choice of AI assistant. The platform includes GLM 5.3 Flash and GLM 5.3 as primary options, with Kimi K3 and DeepSeek v4 Pro available as experimental choices. Usage is tracked as a single monthly allocation with no daily limits, though there is a hard stop at 100 percent of the monthly quota. Premium modes remain private and untouched by the data-sharing arrangement.

The research preview runs from September 14 through October 14, 2026, giving developers a one-month window to test Forge and contribute to the open-model training effort. The limited timeframe allows Bolt.new and Arcee AI to gather feedback and refine the program before a potential broader rollout.

Why Does This Matter for the Broader AI Ecosystem?

Forge represents a deliberate bet on open-source models as an alternative to the proprietary AI services that currently dominate the coding-assistance market. By making open models more accessible and capable, Bolt.new and Arcee AI are attempting to shift the economics of AI development away from vendor lock-in toward a model where capability is distributed rather than rented. This approach directly addresses a key criticism of current AI coding tools: that their cost and reliance on proprietary models limit who can participate in building software.

The partnership also signals confidence in open-weight models as a viable path forward. Arcee AI's Trinity model family is designed for real-world deployment across devices, on-premises infrastructure, and cloud environments, giving developers the flexibility to inspect, adapt, self-host, and deploy AI without dependence on a single proprietary provider. For developers concerned about vendor control or data privacy, this represents a meaningful alternative to closed systems.

"With Forge, you forge your own products and businesses. And by building, you're helping the world forge capable open-weight models. A year ago, a lot of these builders didn't have the purchasing power to do this at all. Arming them is the whole point," said Eric Simons, co-founder and CEO of Bolt.new.

Eric Simons, co-founder and CEO of Bolt.new

The initiative also reflects a broader industry shift toward democratizing AI capabilities. As frontier AI models become increasingly expensive to run and access, open-source alternatives trained on real-world usage data offer a path for smaller teams and individual developers to build competitive applications without massive budgets. Forge's explicit focus on widening access suggests that cost and availability, not just capability, are becoming central competitive factors in the AI development tools market.