
What Are Spur AI Credits?
Spur AI Credits plug straight into a coding harness โ the kind of AI-powered environment you can see in the screenshot above. It looks and feels like a Cursor-style setup: an assistant that lives inside your editor, writes and edits code alongside you, and runs through VS Code, powered by a tool like Voyager Code. It also works with Claude Code โ many builders have used Claude Code to get it set up in minutes.
Every K-Dub Hacks participant gets $2,000 in free AI credits from Spur Innovation. They are simple: a free usage allowance for the AI models you build with all weekend. The interesting part is what one credit buys โ and that is where the open-weights story comes in.
The short version
A Spur AI Credit is a unit of AI usage, metered against the API calls you make from your editor, agent, or app. You spend it the way you spend a cloud credit: request by request, token by token.
At K-Dub Hacks, each of the first 200 sign-ups gets $2,000 in credits, so teams can build with AI without putting a card on file. Claim at check-in, spend it over the weekend.
The original allocation covered 100 participants. We hit that mark far faster than expected โ the response was so strong that the event has decided to double the pool and extend it to 200 participants.
๐What's the benefit?
Free runway to build
You get $2,000 of AI usage without touching a personal card. That is the difference between a weekend prototype and something you can actually finish.
Speed without a premium tier
Open-weight models stream fast as standard โ no 2x-priced "fast mode" required. Your credits go further per minute of work, too.
Experiment freely
Try the agent loop, discard it, try another. Credits are meant to be burned on iteration, not hoarded for the final demo.
Auditable and portable
The open-weight tier ships MIT-licensed weights you can inspect, fine-tune, and run yourself โ no silent model rerouting underneath you.
๐ Frontier-quality credits, proven under a deadline
These are not toy credits. They are the same class of frontier-quality models professional developers use every day, and they plug into the tools you already work in โ VS Code, the most widely used IDE on the market, plus modern agent harnesses and API clients.
K-Dub Hacks co-host Corey Barron is a three-time hackathon winner โ three wins in a row, across three consecutive weeks.
In one of those wins, he used Spur's GLM 5.3 to build a mobile app and update and integrate the accompanying website โ and shipped the entire winning build inside a single 8-hour hackathon session.
If a model can carry a winning build from first commit to demo in eight hours, imagine what $2,000 of it can do over a 48-hour weekend.
๐ Why a credit goes further than you think
A credit is only worth its exchange rate. Our position paper The Open-Weights Advantage priced the same work across frontier and open-weight models โ and the multiplier is the whole story. Put simply: if a session costs $6.00 on one tier and $0.11 on another, your $2,000 buys roughly 330 sessions or roughly 18,000. That is not a discount. It is a different economy.
Open-weight models average 98-99% of the frontier scores across shared coding and agentic benchmarks โ and win outright on security and automation evals.
A standard 500K-in / 50K-out coding session costs about $6.00 on a frontier API versus $0.11 on an open-weight model. With prompt caching, closer to 150x.
The open tier streams around 169 tok/s with a ~1.2s time-to-first-token, versus ~53 tok/s on the flagship โ whose 2x-priced fast mode only reaches 86.
MIT-licensed weights you can download, hash, audit, fine-tune, and air-gap. What you tested is what answers, every time.
The paper's recommendation is the same policy we suggest for your weekend: default open, escalate rare, audit everything.
- Volume and agent loops โ drafts, scaffolding, tests, docs, ETL, chat โ run on the cheap open tier.
- Hard coding and security โ refactors, long builds, vuln discovery โ use the strong open model that takes state-of-the-art on CyberGym.
- The weird 1-5% โ novel multimodal or extreme long-horizon work โ escalate to a flagship, per task, with a budget cap.
Even reserving 5% of calls for frontier escalation, a team's blended cost lands roughly 3-20x below an all-flagship default โ with comparable output quality.
๐ฅ Watch
What are Spur AI Credits โ and why do they matter?
How to set up your Spur AI Credits
๐งญ How to use them
Register for K-Dub Hacks
Credits are reserved for the first 200 sign-ups and handed out at check-in. Register early to lock yours in.
Claim your credits
Log into the participant portal and open the AI Credits page to see your allocation and claim instructions.
Point your tools at the API
Plug your key into your editor, agent harness, or API client. The setup video below walks through it end to end.
Build all weekend
Drafts, tests, refactors, docs, data pipelines โ let the credits absorb the token bill while you ship.
Claim your credits
If you haven't signed up yet, join on Luma to secure your spot among the first 200 participants and claim your $2,000 in Spur AI Credits.