Poppy gives you several AI models. Pick the model under the chat box. Bigger models cost more credits per message.
Chat models
One credit is about 1 cent of processing cost. A message costs credits for the text the model reads (in) and the text it writes (out), so cost depends on how much you have connected and how long the chat is. The numbers below are credits per 1 million tokens.
Model | Best for | Credits per 1M tokens (in / out) | Token limit | Thinking levels |
Claude 5 Sonnet | Ideation and scripting. The default model. | 200 / 1,000 | 200k | Instant, Low, Medium, High |
GPT-5.6 Terra | Frontier reasoning for complex problems. | 250 / 1,500 | 1 million | Instant to Extra High |
GPT-5.5 | Advanced reasoning and problem-solving. | 500 / 3,000 | 1 million | Instant to Extra High |
Claude Opus 5.5 | Most capable model for complex reasoning and agentic coding. | 400 / 2,000 | 1 million | Instant to Extra High |
Claude Opus 5 | Previous-generation Opus. | 500 / 2,500 | 1 million | Instant to Extra High |
Claude Opus 4.8 | Complex reasoning and agentic coding. | 500 / 2,500 | 1 million | Instant to Extra High |
Claude Opus 4.6 | Previous-generation Opus for creative projects. | 500 / 2,500 | 1 million | Instant to Extra High |
Gemini 3.8 Flash | Fast, great for long videos and large datasets. | 75 / 375 | 1 million | None |
GPT-5.4 Mini | Budget-friendly daily content. The cheapest option. | 75 / 450 | 400k | Instant to Extra High |
DeepSeek V4.1 Flash | Very low cost. Reads images too. | 30 / 120 | 1 million | None |
DeepSeek V4 Pro | DeepSeek's strongest reasoning model. Text only, it can't read images. | 132 / 396 | 1 million | None |
Grok 4.7 | Newest Grok. Trending topics and current events. | 200 / 600 | 500k | None |
Grok 4.5 | Trending topics and current events. | 200 / 600 | 500k | None |
Claude Opus models only show for Power User and AppSumo plans. If you do not see them, your plan does not include them.
Gemini 2.5 Pro and GPT-4o are no longer in the list. Chats that used them switch to Gemini 3.8 Flash and GPT-5.4 Mini the next time you open the board.
Image models
Model | Best for | Credits per image (about) |
GPT Image 2.5 Flare | Crisp text and layouts. The default image model. | Low 0.6, Medium 1.4, High 5.3, plus about 1 per reference image |
GPT Image 2.5 Sunburst | Newest OpenAI image model. | Same as Flare |
NanoBanana 2.1 | Fast everyday image model. | 4 at 1K, 6 at 2K |
NanoBanana Pro | Rich detail and lighting. | 15 at 1K or 2K, 30 at 4K |
Seedream 4.5 is in testing and only shows for some accounts. For how to make images, see Image Generation in Poppy.
What are token limits?
Tokens are the amount of text a model can read at once. These count toward the limit:
Everything connected to the chat (documents, videos, links)
The conversation history
A long chat can hit the limit even with nothing connected.
When you hit the overload limit
If you see "Yikes! Info Overload", you have connected more content than the model can read. See Troubleshooting Overload Limits.
How to switch models
Click the model name under the text box.
Pick a model. Click More Models to see the rest.
To change how long the AI thinks, click Thinking and pick a level. See Understanding Thinking Modes.
To compare models, send the same prompt, switch model and send it again.