Silk Compute Plans and pricing

Every plan covers more usage than it costs, at the same per-token prices as credits.

Go

For chatting and light agent use.

$9/month

Billed $100 a year, 2 months free

$12 of usage a month

1.2x

≈ 9,500 agent requests a month

  • All models, in the chat and the API
  • Cancel anytime
Choose Go

Pro

For side projects and your first agent.

$17/month

Billed $200 a year, 2 months free

$25 of usage a month

1.25x

≈ 20,000 agent requests a month

  • All models, in the chat and the API
  • Cancel anytime
Choose Pro

Max

Most popular

For daily work in Claude Code, Codex or OpenCode.

$42/month

Billed $500 a year, 2 months free

$70 of usage a month

1.4x

≈ 55,000 agent requests a month

  • All models, in the chat and the API
  • 2.8x the usage of Pro
  • +5% on credits you buy
Choose Max

Ultra

For agents that run all day, and for teams.

$167/month

Billed $2,000 a year, 2 months free

$300 of usage a month

1.5x

≈ 240,000 agent requests a month

  • All models, in the chat and the API
  • Priority support
  • +20% on credits you buy
Choose Ultra

Each square is $1 of usage a month. Gray squares show the usage you pay for. Light gray squares show the extra usage included.

Enterprise

One contract, one invoice, and a named engineer who rolls it out with your team.

Talk to us
  • SAML SSO and SCIM
  • Roles and model access
  • IP allowlists
  • Audit log streaming
  • Dedicated capacity
  • Custom rate limits
  • Service-level agreements
  • Annual contracts and invoicing
  • A shared Slack channel

Compare plans

GoProMaxUltraEnterprise
Usage
Price a month, billed yearly$9$17$42$167Custom
Usage a month$12$25$70$300Custom
Usage for the monthly price1.2x1.25x1.4x1.5xCustom
Extra on credits you buyNot includedNot included+5%+20%Custom
All models, in the chat and the APIIncludedIncludedIncludedIncludedIncluded
Your team
One account for your whole teamIncludedIncludedIncludedIncludedIncluded
A budget on every API keyIncludedIncludedIncludedIncludedIncluded
Spend by person, key, and modelIncludedIncludedIncludedIncludedIncluded
Activity log and CSV exportsIncludedIncludedIncludedIncludedIncluded
Priority supportNot includedNot includedNot includedIncludedIncluded
Control and security
Your rules on every requestNot includedNot includedNot includedNot includedIncluded
SAML SSO and SCIMNot includedNot includedNot includedNot includedIncluded
Roles and model accessNot includedNot includedNot includedNot includedIncluded
IP allowlistsNot includedNot includedNot includedNot includedIncluded
Audit log streamingNot includedNot includedNot includedNot includedIncluded
Service
Dedicated capacityNot includedNot includedNot includedNot includedIncluded
Custom rate limitsNot includedNot includedNot includedNot includedIncluded
Service-level agreementsNot includedNot includedNot includedNot includedIncluded
A named engineer and a shared Slack channelNot includedNot includedNot includedNot includedIncluded
Paying
Card or walletIncludedIncludedIncludedIncludedIncluded
Crypto, paid up frontIncludedIncludedIncludedIncludedIncluded
Annual contracts and invoicingNot includedNot includedNot includedNot includedIncluded

Pay as you go

Buy credits once, from $5, by card or crypto. No renewal, and they never expire.

Model prices

Per 1M tokens. Plan usage and bought credits are spent at these prices.

ModelModalitiesInput $/MCached $/MOutput $/M
Kimi K3.1 ProBest at cyberNewkimi-k3.1-pro-uncensored$3.50$0.35$8.00
DeepSeek V4.1 Flash 1017deepseek-v4.1-flash-1017-uncensored$0.24$0.038$0.90
DeepSeek V4.1 Flash 1017 Fastdeepseek-v4.1-flash-1017-fast-uncensored$0.24$0.05$0.90
GLM-5.3glm-5.3-uncensored$2.00$0.30$4.00
GLM-5.3 Flashglm-5.3-flash-uncensored$0.24$0.06$1.00
Hy4hy4-uncensored$1.60$0.30$3.20

Pricing questions

Something else? Contact us.

How do the plans work?

Each plan covers your usage up to an allowance a month: $12 on Go, $25 on Pro, $70 on Max, $300 on Ultra, spent at the per-token prices on this page. It resets every month; past it, requests use your credits.

Does unused usage roll over?

A plan's usage doesn't: what a month leaves unused is gone when it resets. Credits you buy never expire, and are only spent once the plan's usage runs out.

Is there a message cap on a plan?

No. A plan sets how much usage it covers, not a number of messages. Past it, an agent keeps running on credits, and auto top-up can refill them.

Can I change or cancel my plan?

Yes, from the Billing page. An upgrade is charged right away for the rest of the period and adds its extra usage for that time; the full allowance starts at the next reset. A downgrade starts at your next renewal, and you keep the plan you paid for until then. Cancelling stops the renewal at the end of the period you paid for, and your credits stay.

Do plans change what credits cost?

From Max up, they add to them. While the plan is active, every credit purchase, by card, crypto or auto top-up, gets extra credit: +5% on Max, +20% on Ultra.

Is paying yearly cheaper?

Yes: 2 months free on every plan when paid yearly.

Can I pay for a plan with crypto?

Yes, up front: a month or a year of a plan, with no KYC, in Monero, Bitcoin, Ethereum and stablecoins. It doesn't renew; paying again for the same plan extends it. Plans paid by card or wallet renew through Stripe.

Can my team share a plan?

Yes. Invite people to your account and they spend from its plan and credits with their own API keys. Each key can have a budget and an expiry date, and you can see spend by person, by key, and by model.

What does Enterprise add?

Rules on every request, SAML SSO and SCIM, roles and model access, IP allowlists, audit log streaming, dedicated capacity, custom rate limits, service-level agreements, and annual contracts with invoicing. It starts with a conversation, so talk to us.

What does "agent requests" mean?

An estimate of the requests a coding agent such as Claude Code, Codex or OpenCode makes on DeepSeek V4.1 Flash 1017. Agents resend their context every turn, so most input is cached and billed at the cached rate. We assume 97% cached input, 2% uncached input and 1% output.