Same models. 4–5× the tokens per dollar.
Per-token rates at ~40 % below OpenRouter list price, and plans whose allowance multiplies it again. Three plans or pure pay-as-you-go — card via Stripe or USDT / USDC stable-coins, no KYC.
Prices in USD — billed by card or crypto.
Free
No card. Resets every 7 days.
- ✓Up to 3.9M tokens / week on DeepSeek-V4-Flash
- ✓10 requests / min
- ✓DeepSeek-V4-Flash, Kimi-K2.6 and MiniMax-M2.7
- ✓Signup bonus on the prepaid balance
Mini
For solo devs and side projects.
- ✓Up to 220M tokens / mo on DeepSeek-V4-Flash (~78M on MiniMax-M2.7, ~39M on Kimi-K2.6)
- ✓≈4× what $9 buys at OpenRouter list price
- ✓Backup provider when the network is busy — paid plans only
- ✓30 requests / min
- ✓Overflow billed pay-as-you-go from prepaid balance
Standard
Most popularDaily-driver plan for active vibecoders and agents.
- ✓Up to 790M tokens / mo on DeepSeek-V4-Flash (~275M on MiniMax-M2.7, ~139M on Kimi-K2.6)
- ✓≈4.6× what $29 buys at OpenRouter list price
- ✓Backup provider + GLM-5.2, the longest context in the catalogue (1M tokens)
- ✓100 requests / min — 10× the free limit
- ✓Tool-calling & streaming included
Team
Small studios and agencies running prod agents.
- ✓Up to 2.3B tokens / mo / seat on DeepSeek-V4-Flash (~820M on MiniMax-M2.7, min 2 seats)
- ✓≈5× what $79 buys at OpenRouter list price
- ✓300 requests / min — shared org billing
- ✓Webhook usage logs + analytics
How the allowance works
Your plan holds a usage allowance, not a token counter. Every request spends it at that model's published per-token price — so a cheap model goes further, an expensive one goes less far, and you never have to predict your input/output split in advance.
Illustrative — your console shows the live figures.
- ✓ Run out and nothing breaks. Overflow bills pay-as-you-go from your prepaid balance at the rates below. Requests do not start failing.
- ✓ Model burn rate is published. DeepSeek-V4-Flash spends the allowance at ×1, MiniMax-M2.7 at about ×2.9, GLM-5.2 at ×5.4 and Kimi-K2.6 at about ×5.7 — shown next to the model in the console, and derived from the same price table on this page.
- ✓ Nothing is stranded. One pooled allowance, not separate input and output buckets, so a prompt-heavy workload can't leave half the plan unused.
The token figures on the plan cards are the allowance converted at today's prices for the model named on each card. The allowance is the limit; the token count is what it currently buys.
Per-token pricing (USD per 1M tokens)
Overflow above plan quota, or pay-as-you-go without a plan. Top up your balance, charged per request.
| Model | Input / 1M | Output / 1M |
|---|---|---|
| DeepSeek-V4-Flash Plan sizes quoted in this model | $0.084 | $0.168 |
| Kimi-K2.6 | $0.35 | $1.46 |
| MiniMax-M2.7 | $0.18 | $0.72 |
| GLM-5.2 Paid plans only | $0.38 | $1.19 |
Roughly 40% below OpenRouter list price for the same models. Billed in USD by card or crypto. Prices verified August 14, 2026.
What 1M tokens costs across providers
Vendor list prices, blended for a 4:1 input-to-output workload — the ratio our traffic actually runs at. Comparing input prices alone flatters whichever model has the cheaper prompt tier, so we don't.
| Model | Input / 1M | Output / 1M | Blended (4:1) | vs Kimi-K2.6 |
|---|---|---|---|---|
| DeepSeek-V4-Flashours | $0.084 | $0.168 | $0.101 | — |
| MiniMax-M2.7ours | $0.18 | $0.72 | $0.288 | — |
| Kimi-K2.6ours | $0.35 | $1.46 | $0.572 | — |
| GLM-5.2-FP8ours · paid plans | $0.38 | $1.19 | $0.542 | — |
| Claude Haiku 4.5 | $1.00 | $5.00 | $1.80 | 3.1× |
| Claude Sonnet 5 | $2.00 | $10.00 | $3.60 | 6.3× |
| GPT-5.2 | $1.75 | $14.00 | $4.20 | 7.3× |
| GPT-5.4 | $2.50 | $15.00 | $5.00 | 8.7× |
| Claude Opus 5 | $5.00 | $25.00 | $9.00 | 15.7× |
| GPT-5.5 | $5.00 | $30.00 | $10.00 | 17.5× |
Sources: Anthropic — Claude pricing, OpenRouter — models & pricing. Prices verified August 14, 2026. Full benchmark comparison on /compare/claude-api; per-model detail and burn rates on /models; cost your own volume with the token cost calculator.
Pricing FAQ
▸How can I pay?
By card (Stripe) or crypto (USDT / USDC stable-coins). No KYC — subscribe to a monthly plan or just top up and pay-as-you-go.
▸How many tokens do I actually get?
Each plan carries a usage allowance, and a request spends the allowance at that model's per-token price. On DeepSeek-V4-Flash the Mini plan works out to roughly 220M tokens a month, Standard to 790M, Team to 2.3B per seat. Other models cost more per token and go proportionally less far — MiniMax-M2.7 about a third as far, Kimi-K2.6 about a fifth. Because the allowance is priced, not counted, you are never forced to guess your input/output split in advance — and nothing is stranded if you guess wrong.
▸What happens if I exceed my plan allowance?
Overflow is billed pay-as-you-go from your prepaid balance at the per-token rates below. No surprise denial — your agents keep running as long as there is balance.
▸Is there a rate limit on plans?
Three rolling windows — 5 hours, one week, and the billing period — each shown as a percentage in your console, the same model Claude uses. They spread the allowance so one burst can't drain the month. If you hit a window with no balance, requests pause until it rolls over (minutes for the 5-hour one), or you can top up to keep going instantly. Most users never notice it.
▸Why show percentages instead of a token counter?
Because a token counter is only meaningful if you also know the model, the input/output split and the price — three things that move. The percentage answers the question you actually have: how much of my plan is left, and when does it reset. The per-token rates are published below if you want to do the arithmetic yourself.
▸Do prices fluctuate?
USD prices are fixed and predictable. Crypto top-ups credit your USD balance at the current rate, so your effective bill stays stable.