Beta
中文 Start→
Closed beta: top-ups come with 30% extra credits, subscriptions with 50–70% extra. We welcome your feedback on the support page.

Pricing

Quoted first.
Charged after.

Priced on the result. A free estimate comes first — a static scan that takes seconds and runs none of your code. Only continuing takes the 2-credit analysis deposit, which is credited in full against the patch. You decide whether to buy only after the speedup is confirmed end to end in your own program.

Trial pack
$3
10 credits · $0.30 / credit
  • The best rate we offer, once per account
  • ≈ 2–3 full runs on your own repository (a run takes about 2–5 credits)
  • Credits valid 24 months
  • Private code supported (zip upload)
Or pay nothing up front: your first run (on an RTX 4090) requires no deposit, so you can decide after seeing the measured speedup
Usage pack
$5 +
$0.77 / credit · +30% in beta · any amount
  • Credits valid for 24 months
  • Private repos, results kept private
  • Buy more as needed, no commitment
  • A run takes about 2–5 credits ≈ $1.5–4, depending on code size, complexity, what the run finds and the speedup
  • Card or WeChat Pay, any amount; WeChat Pay requires a QR scan on desktop
Pro +50% credits · beta
$28 / mo
42 credits · $0.67 / credit
  • 50% bonus credits per dollar during the beta
  • ≈ 10–15 full runs per month
  • Private repos · 5 concurrent
Max +60% credits · beta
$99 / mo
158.4 credits · $0.63 / credit
  • ≈ 40–60 full runs per month
  • Private repos · priority queue · 5 concurrent
Max+ +70% · beta
$199 / mo
338.3 credits · $0.59 / credit
  • ≈ 85–130 full runs per month
  • Top queue priority · 5 concurrent
  • For heavy individual use: large repos, many parallel runs
Beta bonus · top-ups +30% · Pro +50% · Max +60% · Max+ +70% Subscription credits reset each period · packs last 24 months Packs: WeChat Pay (desktop QR) or card · Subscriptions: card only · WeChat users can buy the 30-day pack at the same price and bonus

What a run costs

Compute always at 50%, tokens at 8–13% by speedup

Metered on the compute and tokens actually consumed, against public list prices: GPUs at Beam's or Modal's list price, tokens at Anthropic's published API rates. The GPUs are hardware we rent, and they cost the same whatever the run finds, so compute is always billed at 50%. A run that never uses a GPU (submitted as CPU only) has no machine charge at all. Tokens are the agent's work, and the speedup determines what that work is worth. Below is a real RTX 4090 run: 2.8 hours, 1.61× end to end, whose tokens land in the 11% tier.

ComputeGPU + sandbox, by the second
5.5 × 50% = 2.8
Model tokensreading, editing, verifying
12.7 × 11% = 1.4
At list ≈ 18.2 creditsYou pay ≈ 4.2 credits
listyou pay
8%speedup under 20%
9%20%–50%
11%50%–100%
13%100% and up

The share of list you pay for tokens depends on the end-to-end speedup the run delivered — the run above reached 1.61x, the highlighted tier. That speedup is the headline number in your report, and you can verify it yourself. Below 1.10x the patch is not sold: the deposit is returned and nothing is charged.

When you are charged

1 · Free estimate
0
A static scan that takes seconds; none of your code runs and no GPU is used. You get a projected speedup and its range, then decide.
→
2 · Continue
2 credits
The analysis deposit covers reading the repo, building the harness and timing the baseline. Credited in full against the patch; returned when no speedup is found.
→
3 · Buy the patch
quote − 2
Only after the speedup is confirmed end to end in your own program. The quote reflects actual usage, and you never pay more than the quote.
Target missed (< 1.10×)No patch is sold, you still get the report, and the deposit is returned.
The failure is oursInfrastructure failure, timeout, redeploy interruption, an environment that fails to build, an entry point whose running code we cannot locate, or a program too large for our biggest GPU — the deposit is returned.
Estimate skippedThe deposit is taken at submit; cancelling within 5 minutes still returns it.
SubscribersPro waives 1 deposit per period; Max / Max+ waive 3. Stopping at the estimate is always free.

How quickly 4.2 credits pay for themselves

You buy the patch once; the speedup lasts. At 1.61×, the same work takes 62% of the machine time, saving 38% of your compute cost every hour it runs. At public cloud rates:

RTX 4090$0.7 / h · saves $0.27
16 hours to pay back
RTX 5090$1.1 / h · saves $0.42
10 hours
H100$6.9 / h (AWS) · saves $2.61
2 hours

For a job that retrains 20 hours a week on an H100, the first month's savings alone far exceed the cost of the patch.

Billing rules

When you are charged, and when you are refunded

EstimateFree; none of your code runs.
Continue2-credit deposit, credited in full against the patch.
UnlockCharged at the quote, which is actual usage with no markup; never more.
Target missedNo patch sold; you still get the report; deposit returned.
Our failureThe deposit returns to your balance.
Deduction orderCurrent-period subscription credits first, then packs, oldest first.

Speedup = the end-to-end wall-clock time of your own command, optimised versus the frozen baseline. The report breaks it down into startup, the key loop and everything after it.

See the numbers first. Then decide.

Questions? Email us