Models and rates

The models behind the Agent and the AI steps in loops, and what each one costs per million tokens.

Pluto 1

Pluto 1 is Pluto's model built for ads. It's the default in every workspace and comes first in the model picker. When a new Pluto version ships, Pluto in the picker moves to it. A version that's replaced gets a deprecation date, with a notice before it.

Pluto 1 thinks at Medium effort unless you pick High or Extra high beside it in the picker. Higher effort takes longer to answer and uses more credits.

4 November 2026: Medium is now Pluto 1's default, so answers come faster. It was High. Chats and loop steps where you picked High keep it.

How usage is billed

AI runs on credits: 1 credit = $0.01. Every plan includes credits each month, shared by your organization; the free trial includes 300 for the whole trial. Unused credits roll over for 12 months. On every paid plan, owners and admins can buy more at any time, paid by card before you use them. See Billing.

Usage is counted in tokens, the pieces of text a model reads and writes. Each request uses credits worth its charge at its model's rates:

  • Input: what the model reads: your message, the chat so far and what the Agent looked up. Images the Agent looks at, such as your media and your ads' creatives, count as input too.
  • Cache write and Cache read: input the model keeps for the next steps of the same task. Reading it back costs much less than reading it fresh.
  • Output: what the model writes, including its reasoning.

Rates below are per million tokens, in US dollars, and exclude applicable taxes. A rate of $6.00 is 600 credits.

Pluto

ModelInputCache writeCache readOutput
Pluto 1$6.00$7.50$0.30$30.00

Web searches: $15.00 per 1,000.

Anthropic

ModelInputCache writeCache readOutput
Claude Opus 5.5$5.28$6.60$0.264$26.40
Claude Sonnet 5.5$2.64$3.30$0.264$13.20
Claude Opus 5$6.60$8.25$0.66$33.00
Claude Sonnet 5$2.64$3.30$0.264$13.20
Claude Haiku 4.5$1.20$1.50$0.12$6.00

Web searches: $12.00 per 1,000.

OpenAI

ModelInputCache writeCache readOutput
GPT-6 Astra$13.20$16.50$1.32$66.00
GPT-6.1 Sol$2.64$3.30$0.132$13.20
GPT-6 Sol$2.64$3.30$0.264$13.20
GPT-6 Luna$0.132$0.165$0.0132$0.66
GPT-5.6 Sol$5.28$6.60$0.528$26.40
GPT-5.6 Terra$2.64$3.30$0.264$15.84
GPT-5.6 Luna$0.264$0.33$0.0264$1.584
GPT-5.5$6.60$6.60$0.66$39.60
GPT-5.4$3.30$3.30$0.33$19.80
GPT-5.4 Mini$0.99$0.99$0.099$5.94
GPT-5.2$2.10$2.10$0.21$16.80
GPT-4.1$2.40$2.40$0.60$9.60
GPT-4.1 Mini$0.48$0.48$0.12$1.92
GPT-4o$3.00$3.00$1.50$12.00
GPT-4o Mini$0.18$0.18$0.09$0.72

Requests over the long-context threshold are charged at these rates for the whole request:

  • GPT-6 Astra: over 272K tokens, $26.40 input and $99.00 output.
  • GPT-6.1 Sol: over 272K tokens, $5.28 input and $19.80 output.
  • GPT-6 Sol: over 272K tokens, $5.28 input and $19.80 output.
  • GPT-6 Luna: over 272K tokens, $0.264 input and $0.99 output.
  • GPT-5.6 Sol: over 272K tokens, $10.56 input and $39.60 output.
  • GPT-5.6 Terra: over 272K tokens, $5.28 input and $23.76 output.
  • GPT-5.6 Luna: over 272K tokens, $0.528 input and $2.376 output.
  • GPT-5.5: over 272K tokens, $13.20 input and $59.40 output.
  • GPT-5.4: over 272K tokens, $6.60 input and $29.70 output.

Web searches: $12.00 per 1,000.

xAI

ModelInputCache writeCache readOutput
Grok 4.7$2.64$2.64$0.66$7.92
Grok 4.6$2.64$2.64$0.66$7.92
Grok 4.5$2.40$2.40$0.36$7.20
Grok 4.3$1.50$1.50$0.24$3.00

Requests over the long-context threshold are charged at these rates for the whole request:

  • Grok 4.7: over 199,999 tokens, $5.28 input and $15.84 output.
  • Grok 4.6: over 199,999 tokens, $5.28 input and $15.84 output.
  • Grok 4.5: over 199,999 tokens, $4.80 input and $14.40 output.
  • Grok 4.3: over 199,999 tokens, $3.00 input and $6.00 output.

Web searches: $6.00 per 1,000.

Choose a model

Pick a model in the chat box as you write, or Auto to have one chosen for you: Claude Sonnet 5.5, moving to Claude Opus 5.5 for large plans.

Models under My account in Settings decides which models your picker shows, the same in every workspace. Pluto 1 is always shown. To start, the picker shows Pluto 1, Claude Opus 5.5, Claude Sonnet 5.5, GPT-6 Astra, GPT-6.1 Sol, GPT-6 Luna and Grok 4.7. New models start hidden.

In SettingsAgent (Settings, then Agent), under Workspace, the workspace's owners and admins choose which models are on under Models. Your picker shows only models the workspace allows.

Good to know

  • Requests from the Agent and Loops pass through a gateway in the European Union (Sweden). The model providers process them in the United States and other countries where they operate, so AI requests aren't limited to processing in the US. See the Privacy Policy.
  • Your organization's Usage page in Settings shows every request with its model and credits, and lists these rates under Models and rates. Pluto 1's requests show credits rather than tokens.
  • Creative Memory comes with Enterprise and reads new ads with your organization's credits. It isn't counted in tokens: reading uses a set number of credits per image and per minute of video. See Media library.
  • Using Pluto Ads from your own AI tools through MCP or the API is included on every plan and never uses credits.