Pay only for the tokens you use.
Prepaid wallet. Zero tiers. Zero subscriptions. Top up $5, spend it on whichever model you want, top up again when you run out. We add a flat 7% to whatever the upstream provider charges. That's it.
Cents-level negative balance on your final request before lockout — same as OpenAI's prepaid model. Top up to unblock.
Sample model prices
A few entries from our catalog. Live prices live in your dashboard. Prices below are what you pay — upstream cost × 1.07.
| Model | Provider | Context | Input / 1M | Output / 1M |
|---|---|---|---|---|
DeepSeek V3 deepseek-v3 | deepseek | 64K | $0.289 | $1.18 |
DeepSeek R1 (reasoning) deepseek-r1 | deepseek | 64K | $0.589 | $2.34 |
Llama 3.3 70B Instruct llama-3.3-70b | groq | 128K | $0.631 | $0.845 |
Qwen3 Coder 480B qwen3-coder-480b | cerebras | 128K | $0.642 | $1.28 |
Claude 3.5 Sonnet (passthrough) claude-3-5-sonnet-20241022 | anthropic | 200K | $3.21 | $16.05 |
Cached identical prompts from Cloudflare AI Gateway are $0 regardless of provider — they don't hit the upstream.
What every account gets
- Anthropic
/v1/messages+ OpenAI/v1/chat/completions— point any SDK at one base URL - 12+ providers / 70+ models, growing weekly
- :nitro / :floor routing + sequential fallback on 5xx
- Live edge cache — duplicate prompts return for $0
- Real-time wallet balance + per-request usage in your dashboard
- Reasoning streams (DeepSeek-R1, etc.) translated into Anthropic
thinkingblocks - SOC 2-grade isolation via Cloudflare Workers + Durable Objects
- No subscription, no overage bill, no salesperson
$5 starter credit on signup. No credit card required.