How credits are consumed in windsurf

In Windsurf, every AI request consumes credits, but the way credits are deducted depends on the model you’re using. There are two main billing modes:

1. Flat per-prompt cost

Some models use a fixed number of credits per request. For example:

• SWE-1 Lite → 0 credits (free to use on all plans)

• SWE-1 → 0 credits (included on Pro and above)

• GPT-5 Low Reasoning → 0.5 credits per prompt

• GPT-5 High Reasoning → 1.5 credits per prompt

With this method, the length of your prompt or output doesn’t change the cost — you pay the same credit amount every time you hit “send.”

2. API pricing based on tokens

Other models charge credits based on the number of tokens processed, similar to how providers like OpenAI or Anthropic bill directly.

• Windsurf uses the model provider’s API price + 20% margin

• Converts the dollar amount to credits at $0.04 per credit

• Bills separately for input tokens, cache-read tokens, and output tokens

Example: Claude Sonnet 4 in Windsurf:

• Input: 90 credits per 1M tokens = $3.60 per 1M input tokens

• Cache read: 9 credits per 1M tokens = $0.36 per 1M cached tokens

• Output: 450 credits per 1M tokens = $18.00 per 1M output tokens

Fractional usage is rounded up to the nearest hundredth of a credit. This means a long conversation or high-output model can consume more credits than you expect if you’re not tracking token counts.

The BYOK vs. Windsurf metered divide

Windsurf supports two distinct ways of running models:

• Windsurf-metered usage: Credits are deducted for each request, either via flat-per-prompt or API token pricing.

• Bring Your Own Key (BYOK): Your usage is billed directly by the model provider, and Windsurf does not deduct credits.

How BYOK works

If you’re on a Free or paid individual plan (Pro), you can add your own API key for certain models (e.g., Claude Sonnet, Claude Opus). When you select a BYOK model in the interface, Windsurf routes requests to your provider account instead of its own infrastructure.
This means:

• No Windsurf credits are consumed

• You pay your provider’s standard API rates directly

• Windsurf still provides the same editor interface and features, but without cost control at their end

Why Teams and Enterprise don’t get BYOK

BYOK isn’t offered for Teams or Enterprise plans. While Windsurf doesn’t explicitly state the reason, the likely factors are:

• Centralized billing: Team-level cost management breaks if each user has a separate provider account

• Support obligations: Windsurf can’t troubleshoot API issues if usage is routed through third-party billing

• Analytics and governance: Admin dashboards lose visibility if usage bypasses Windsurf’s metering

Developer perspectiveIf you’re a solo dev already paying for model usage directly (e.g., an Anthropic or OpenAI account), BYOK can save money compared to buying Windsurf credits. But it also shifts responsibility for cost tracking to you, and you lose features like credit-based budget alerts inside Windsurf.

Understanding overage logic and add-on credits

An overage is when you exhaust your monthly credit allocation and still need to use premium models or API-priced requests. 

And once you exhaust your monthly credits, you’ll have to purchase add-on credits to continue usage. Without them, you’re limited to 0-credit models like SWE-1 Lite.

Add-on credit pricing

• Pro plan: $10 for 250 credits ($0.04 per credit)

• Teams & Enterprise: $40 for 1,000 credits ($0.04 per credit)

The per-credit rate for add-ons is the same as Windsurf’s internal conversion rate for API pricing. The difference lies in your bundled credits:

• On Pro, bundled credits are effectively cheaper (500 credits for $15 = $0.03 each)

• On Teams/Enterprise, bundled credits cost more per unit than on Pro because the plan price also covers admin features, security, and support — not just credits

Auto-refill

Windsurf supports automatic credit refills in increments of $10 (Pro) or $40 (Teams/Enterprise), with a monthly budget cap you can set. This avoids the disruption of hitting 0 credits mid-session, but it also increases the risk of silently overspending if you don’t track usage.

If your workflow is steady and predictable, you can budget for your monthly allocation. But if you’re running large batch jobs, fine-tuning, or heavy reasoning mode sessions, your overage costs can quickly match or exceed your base plan fee.

Model differences that affect your bill

Not all models in Windsurf consume credits the same way. Some are effectively free, others cost a predictable per-prompt rate, and others scale with tokens through API pricing. Understanding these differences is critical,  the same workflow can cost close to zero credits with one model or burn through your allocation rapidly with another.

Zero-credit models

• SWE-1 Lite: Always 0 credits, available on all plans

• SWE-1: 0 credits when included in Pro and above

These models allow continuous usage without touching your credit balance — a practical choice for daily development tasks.

Flat per-prompt models

• GPT-5 Low Reasoning: 0.5 credits/prompt

• GPT-5 Medium Reasoning: 0.5 credits/prompt

• GPT-5 High Reasoning: 1.5 credits/prompt

• Gemini 2.5 Pro: 1 credit/prompt

• Kimi K2: 0.5 credits/prompt

• Qwen3-Coder: 0.5 credits/prompt

Here, prompt length doesn’t matter. You pay the same credit each time you run a request. This makes it easy to predict costs, but it can also make “short” queries feel expensive relative to API pricing.

Token-based API models

• Claude Sonnet 4: Input = 90 credits / 1M tokens, Output = 450 credits / 1M tokens

• Claude Opus 4.1: Higher baseline, 20 credits per prompt for Thinking mode

• Others marked BYOK/API: Token usage billed at provider rate + 20%, converted at $0.04 per credit

These models can swing wildly in cost depending on prompt and output size. A lightweight query might cost fractions of a credit, while a long reasoning-heavy answer could burn through dozens.

“Thinking” models

Models like Claude Sonnet (Thinking) or Claude Opus Thinking introduce extra reasoning tokens. Since reasoning tokens are billed as output, they amplify credit consumption. This is where usage becomes least predictable.

Choosing the right plan for your use case

The best Windsurf plan isn’t about the headline price, it’s about how your team actually uses credits, which models you rely on, and how much governance you need. Here’s how the tiers map to real-world scenarios:

1. Free: for experiments and hobby projects

If you’re just testing Windsurf or running small-scale side projects, the Free plan works. You get 25 credits/month and unlimited SWE-1 Lite, which is enough for light prototyping. BYOK lets you connect your own API key, so if you’re already paying Anthropic or OpenAI directly, you can still use Windsurf’s editor without consuming credits.

2. Pro: for solo developers building regularly

At $15/month, Pro gives you 500 credits (≈ $20 value) plus access to SWE-1 and 5 deploys/day. This plan is designed for individuals running steady workloads. The bundled credits are cheaper per unit than add-ons, so even modest daily use quickly justifies the cost. For predictable development cycles, this is the most cost-effective plan.

3. Teams: for small teams that need control

At $30/user/month, Teams is less about cheaper credits and more about operational features. You still get 500 credits/user, but the value comes from centralized billing, admin analytics, priority support, and optional SSO. If you’re managing multiple devs and don’t want to chase expense reports or API receipts, Teams is the baseline.

4. Enterprise: for large-scale organizations

Starting at $60/user/month, Enterprise adds RBAC, advanced SSO, access controls, hybrid deployment, and dedicated support. Once your org crosses 200 seats, the bundled credits double to 1,000/user (≈ $40 value). This plan makes sense when compliance, governance, and predictable scaling are as important as raw credit cost.

• Use Free if you’re experimenting or relying on BYOK.

• Upgrade to Pro if you’re a solo dev burning credits every day.

• Choose Teams if you’re managing multi-user workloads and need billing visibility.

• Go Enterprise if your org is scaling past 50+ seats, needs SSO/RBAC, or wants hybrid deployment flexibility

Wrapping up

Windsurf’s pricing isn’t complicated once you break it down, but the details matter. The monthly plans set the baseline, bundled credits establish predictable value, and the real variability comes from which models you use and how you consume tokens

Windsurf’s pricing strategy balances predictability for solo developers with control and governance for teams. If you track your usage closely and match the right plan to your workflow, you can avoid most surprises. But if you treat credits as “set and forget,” you risk higher bills than expected, especially with token-heavy models.