Highlights
- Bolt Forge is a new open-model-sourced agent in the Bolt.new platform. It runs on multiple models: GLM 5.3 Flash by default, GLM 5.3 alongside it, and Kimi K3 and DeepSeek v4 Pro as experimental options.
- Every individual Pro plan includes up to 50X more Forge usage at no extra cost through October 14, 2026.
- On the Bolt Build Index, Forge's open models score 92.2 against 101.0 for the top paid model: 91% of the top score.
- Training is always opt-in. Forge sessions are anonymized with secrets stripped, and they help train open-weight models with Arcee AI. The Standard and Max agents never train on your data.
Bolt Forge's open models score 92.2 on the Bolt Build Index, against 101 for the top paid model in Bolt.new. The trade-off is nine points; the payoff is up to 50X more usage at no extra cost on every individual Pro plan.
What is Bolt Forge?
Bolt Forge is an agent in Bolt.new, the AI app builder, that runs on open-source AI models only. It launches on September 14, 2026 as a research preview, and it gives every individual Pro plan up to 50X more usage at no extra cost through October 14. The trade that makes it possible: builders who switch into Forge opt in to share anonymized build sessions, and those sessions help train new open-weight models.
Forge is an agent, not a single model. It sits in the agent picker next to the Standard and Max agents you already use, and those stay private and untouched: Bolt never trains on anything you build outside Forge. Switch in and you're building on the open frontier. Switch back and everything is exactly as it was.
September 14 marks the initial kickoff of the research preview for Forge. After October 14, it continues on as a working lab. The 50X usage, however, is a special offer for all Pro users during this launch preview.
Every Forge session you opt in to helps train the next generation of open-source models, and the allocation is the payment.
How does Bolt Forge work?
Five steps, start to finish:
- Pick Forge in the agent picker. It appears as a third agent next to Standard and Max on every individual Pro plan.
- Opt in with one tap. A consent screen appears each time you switch into Forge and spells out the trade in plain language. Decline it and you stay in your usual agents; nothing changes and nothing is shared.
- Build. Forge usage draws from one monthly bar with no daily limits and a hard stop at 100%. There's no daily math and no surprise pause mid-project.
- Bolt anonymizes shared sessions before anything leaves its infrastructure, stripping secrets, sensitive data, and personal information by default, and validating the pipeline against seeded test data.
- The data goes to work at Arcee AI, a U.S. open-model lab, under a signed data processing agreement, and the models it helps train publish their weights.
Why we created Bolt Forge
Price has decided who gets to build with AI at speed and scale. Builders arrive at Bolt with an idea and the skill to see it through, then ration prompts like fuel: every brainstorm, every rough draft, every dead end burns credits priced for polished work. The drafting phase of building, the part where you need room to wander, is exactly the part usage caps punish.
Forge removes the penalty. The allocation is big enough to draft, test, tear down, and rebuild. Forge is where you test your big idea, play around with new concepts, and experiment as far outside the box as you want to go. Build more for less, and save your premium credits for the work that needs them.
The other reason is the models themselves. AI inference costs have dropped 280x in 18 months (Stanford HAI AI Index, 2025); that collapse is what makes an allocation this size possible. Open-source models also improve when they see how real software gets built, the one thing you can't scrape. Opted-in Forge sessions provide it, with consent and receipts.
Builders get room, and open models get better.
Key components of Forge
The open-source model lineup
Forge runs open models you can see and select. As of September 2026 that means GLM 5.3 Flash by default, GLM 5.3 alongside it, and Kimi K3 and DeepSeek v4 Pro as experimental options. The lineup will evolve, and when a new open model drops, Forge is the first place in Bolt it lands.
Two honest notes. These models are experimental inside Bolt, so duplicate your project before switching a serious build into Forge, and keep complex production work in Standard or Max. And they don't all cost the same to run: Kimi K3 and DeepSeek v4 Pro burn through usage faster than the GLM pair, so start on the default and reach for them when the work calls for it. Forge also can't take PDF uploads yet.
How capable are the open models?
Capable enough that Bolt tested them before shipping them. The Forge lineup ran through the Bolt Build Index, Bolt's own benchmark for how well a model completes real Bolt projects, and came out at 91% of the top paid model's score (Claude Opus 5), as of September 2026.
Dedicated capacity
Bolt runs Forge on its own reserved hardware instead of paying a provider per request. A fixed, predictable cost means more of what you pay goes to building instead of markup.
Powered by WebContainers
Your Forge projects run in your browser on WebContainers, the technology StackBlitz built and Bolt runs on, so there's no server rented for every build. Builds stay fast and costs stay low. Between the reserved hardware and the browser runtime, the cost structure is the reason Forge's price works.
The Forge allocation
Every individual Pro plan includes the Forge allocation at no extra cost through October 14: up to 50X more usage, drawn from a single monthly bar that resets on your renewal date. Every Pro tier gets the same allowance, and Forge usage is separate from your Standard and Max usage. There are no daily limits. When the bar hits 100%, Bolt switches you back to Standard rather than charging an overage.
Consent and anonymization
Training in Bolt Forge is opt-in, every time. Bolt never trains models unless you switch into Forge, and the Standard and Max agents never train on your data at all. A one-tap consent screen fires each time you enter Forge, so the trade stays visible instead of buried in terms you accepted once and forgot. The shared data is your prompts, your code, and the fix traces Bolt creates. Secrets and personal information are stripped and anonymized by default before anything leaves Bolt's infrastructure, and Bolt validates the pipeline against seeded test data.
You can stop whenever you want by switching back to Standard or Max, which ends new sharing; data already used in training stays in the models it helped train. Teams and Enterprise workspaces are excluded from Forge and from all data collection.
The Arcee AI partnership
The training side runs through Arcee AI, the U.S. open-model lab behind the Apache 2.0-licensed Trinity model family. Bolt has partnered with Arcee to help train a trillion-parameter-class open-weight model. Sessions builders share during the preview window, September 14 to October 14, 2026, feed the first training run, which begins in October, and a signed data processing agreement governs every transfer.
What makes the corpus valuable is what's in it. The models learn from the work people do in Forge, and only there: starting projects, making edits, connecting databases, fixing errors. A model that has seen the job gets it right in fewer tries. And the weights publish when the work is done: what builders help train, they can download, inspect, and run, like anyone else.
Benefits of Bolt Forge
What the trade buys you:
- Brainstorms, MVPs, and experiments stop competing with production work for premium credits.
- 91% of the top paid model's Bolt Build Index score, included with Pro at no extra cost.
- Usage you can read at a glance: one monthly bar, no daily limits, a hard stop instead of an overage bill.
- A stake in the outcome: what your builds help train, you can download and run anywhere.
Already on Pro? You have Forge; there's nothing to buy, no line to stand in, and you can start building with Forge now. When the preview window closes on October 14, Forge continues in Bolt as an experimental open-source agent.
How to get started with Bolt Forge
On an individual Pro plan: open the agent picker, choose Bolt Forge, and take the one-tap opt-in. You're building on open models in under a minute.
Not on Pro yet? There are two doors. Join the Bolt Lite waitlist; codes go out in waves, and the first seats open by September 21, 2026. Or skip the line: Pro is $25 a month billed yearly and includes up to 50X Forge usage at no extra cost, no access code required.
Every Forge build does two jobs. It ships your thing, and it teaches models that belong to everyone.
Start building in Forge.
