Open-weight models,
ready to code.
Buildful runs them in the cloud, so there's no GPU or API key to set up, and opens tested pull requests on your repo.
Start buildingFree to try. Then $10 a month, flat.
Open-weight models from
- Moonshot AI
- Z.ai
- DeepSeek
- MiniMax
- Meta
Frontier models from
- Anthropic
- OpenAI
- xAI
- Alibaba
An example task
Open-weight models
did all of this.
Smart Mode picked an open-weight model, which read 142 files, wrote the change and tested it in a browser.
Add a login page with Google sign-in. Keep the existing email login working.
BuildfulWorking…14 minWorking on it…
Added a login page with a Google button and wired the OAuth callback. Email login still works.
- ✓Read 142 files in tidepool/web
- ✓Planned the change
- ✓Wired up the OAuth callback
- ✓Built the sign-in page
- ✓Checks passed: lint, types, tests
- ✓Tested sign-in in a real browserrecordingrecorded
Nothing to set up
Everything you'd have
to set up, done.
Open-weight models with no hardware to buy and no keys to manage.
No GPU.
The models run in the cloud, not on your laptop. Nothing to download or keep running.
No keys.
Smart Mode is included in your plan. Picking a model yourself uses prepaid credit.
More thinking for hard tasks.
Buildful judges how hard each task is and gives the hard ones more time to think.
What it costs
A tenth of
the price.
Open-weight models cost a fraction of frontier models per token. Buildful turns that into one flat price, so you never count tokens.
Price per million output tokens
Coding models, list prices
Open-weight
- GLM 5.3 Flash$0.50
- DeepSeek V4.1 Flash$1.20
- GLM 5.3$4.40
Frontier
- GPT-6.1 Sol$10
- Claude Sonnet 5.5$10
- Claude Opus 5.5$20
- GPT-6 Astra$50
List prices in US dollars. Not every open-weight model is cheap: Kimi K3 is $15. Small frontier models such as Claude Haiku 5.5 are $0.50.
What coding plans cost a month
One price, no token math
- Buildful ProFlat price, no tokens to count$10
- GitHub Copilot Pro$10
- Claude Pro$20
- ChatGPT Plus$20
- Cursor Pro$20
- Claude Maxfrom $100
- Factory Max$200
Other plans at their US prices, from each plan's pricing page on October 5, 2026. Each plan includes different things, so compare what you get back, not only the price.
Who it's for
Developers with
a backlog.
You have a repo and a list of fixes, and you'd rather not sit watching an agent type.
- You like open-weight models, not running them.
- You've tried them on your laptop or with your own keys. Here they run in the cloud, already set up.
- You hit your AI limit by lunch.
- Claude Code needs a paid Claude plan, and Max starts at $100 a month with usage limits. Buildful is $10 a month for 1,000 tasks.
- Your backlog is longer than your day.
- Hand over the bug fixes and dependency upgrades in one go, and review them when you're ready.
- You'd rather review than babysit.
- No terminal to watch. You get a pull request with its checks passed and a browser recording.
How it works
Give it a task.
Come back to a
pull request.
Connect GitHub
Install the Buildful GitHub app and pick the repos it can work on.
Give it a task
A feature or a bug fix, in plain words. Then close the tab.
It builds and tests
In a private cloud sandbox. Your checks run, and UI changes are tested in a real browser.
You review and merge
Watch the recording, read the change, merge.
Checks and tests
Checked before
you see it.
Your checks run
Lint, types and tests, the ones your repo already has.
Failures get fixed
Up to 3 tries, then it reviews its own diff.
UI is tested
In a real browser, with a recording you can watch.
Add CSV export to the orders page
tidepool/web · #218 · 6 files · +184 −12
SUMMARY
Adds an Export CSV button to the orders page. It downloads the orders that match the current filters, with tests for the file format.
- Lint and types
- Tests, after one fix
- Browser test: the file downloads
You don't have to take the model's word for it. Every change shows its checks and its recording.
Models
Smart Mode picks.
Or you do.
Every task in your plan runs on Smart Mode. On Pro and Team you can also pick any model below, paid from prepaid credit.
Included in your plan
Smart Mode
Buildful picks an open-weight model for every task. $10 a month, flat, with no tokens to count.
- ✓ Every plan, Free included
- ✓ Checks and browser test
- ✓ No API key needed
Pick it yourself · prepaid credit
Open-weight
- Moonshot AIKimi K3
- Z.aiGLM 5.3, GLM 5.3 Flash
- DeepSeekDeepSeek V4.1 Flash
- MiniMaxMiniMax M3.1 Flash
- MetaMuse Spark Open
Pick it yourself · prepaid credit
Frontier
- AnthropicClaude Fable 5.1, Claude Opus 5.5, Claude Sonnet 5.5, Claude Haiku 5.5
- OpenAIGPT-6.1 Sol, GPT-6 Astra, GPT-6 Luna
- GoogleGemini Argon, Gemini 3.8 Flash
- xAIGrok 4.7
- MetaMuse Wonder 1.0
- AlibabaQwen 3.8 Max
Frequently asked
questions
Which open-weight models does Buildful use?
Smart Mode picks an open-weight model for you and doesn't name it, and every task in your plan runs on it. On Pro and Team you can pick one yourself, such as Kimi K3, GLM 5.3 or DeepSeek V4.1 Flash, or a frontier model like Claude or GPT, paid from prepaid credit.
Are open-weight models free?
The weights are free to download, but running them isn't: you need a GPU or a paid API. Because the weights are published, many companies compete to host them, and we compared their list prices. With Buildful you pay per task, not per token.
What's the difference between open-weight and open source?
An open-weight model publishes its trained weights, so anyone can run or fine-tune it. Open source in the strict sense also opens the training code and data, which most open-weight models don't release. The Open Source Initiative explains the difference.
Do I need a GPU or my own API key?
No. The models run in the cloud, and Smart Mode is included in your plan. You only add a key if you want a frontier model billed to your own account: Buildful takes keys from OpenAI, Azure OpenAI, Anthropic and xAI.
Is it as good as Claude or GPT?
We don't claim that. What we show is the work: every change runs your checks, failures get up to 3 fixes, and UI changes are tested in a browser and recorded. Read a real task from our own repo. For a hard task you can pick Claude or GPT yourself on Pro.
Is my code private?
Each task runs in its own private sandbox that's deleted when the task ends, and nothing reaches your main branch until you merge. Environment variables you add go into the sandbox so your app can run, so use staging values.
How much does it cost?
Your first 3 tasks are free, with no card. Pro is $10 a month for 1,000 tasks. Team is $40 a month for 3,000 tasks shared by unlimited teammates.
Written by
Sandeep PandaCo-founder, Bug0 · Building Buildful. Previously CTO and co-founder of Hashnode.