Model prices
What each model costs, at list price
18 models from Anthropic, OpenAI and Google, in US dollars per million tokens, as each provider publishes them. Checked on 2026-11-15. Cliff adds nothing: with your own key you pay the provider directly.
Price a prompt
A long agent session sends the conversation again each turn, so most input is cached after the first answer. Cliff prints the real counts after every answer.
The list
The providers' own list prices for standard pay-as-you-go API use. Cliff adds nothing on top: with your own key you pay the provider directly. Prices change; the source links are the final word. Cached input is what a provider charges for the part of a prompt it has seen recently; Anthropic also charges once to write that cache. Cliff turns caching on for Anthropic automatically.
Anthropic
| Model | In | Cached in | Cache write | Out | Notes |
|---|---|---|---|---|---|
Claude Fable 5.2claude-fable-5-2 | $8 | $0.2 | $10 | $40 | |
Claude Opus 5.5claude-opus-5-5 | $4 | $0.2 | $5 | $20 | default |
Claude Sonnet 5.5claude-sonnet-5-5 | $2 | $0.1 | $2.5 | $10 | |
Claude Haiku 5.5claude-haiku-5-5 | $0.1 | $0.01 | $0.125 | $0.5 | Prompts over 100,000 tokens: $0.5 in, $2.5 out. |
Claude Opus 5claude-opus-5 | $5 | $0.5 | $6.25 | $25 | |
Claude Sonnet 5claude-sonnet-5 | $2 | $0.2 | $2.5 | $10 | |
Claude Haiku 4.5claude-haiku-4-5 | $1 | $0.1 | $1.25 | $5 |
Standard API, global routing; cache writes are the 5-minute price. Key: ANTHROPIC_API_KEY. Source: platform.claude.com/docs/en/about-claude/pricing
OpenAI
| Model | In | Cached in | Out | Notes |
|---|---|---|---|---|
GPT-6 Astragpt-6-astra | $10 | $1 | $50 | |
GPT-6.1 Solgpt-6.1-sol | $2 | $0.1 | $10 | default |
GPT-6.1 Terragpt-6.1-terra | $1 | $0.05 | $5 | |
GPT-6 Lunagpt-6-luna | $0.1 | $0.01 | $0.5 |
Standard tier, short context. Key: OPENAI_API_KEY. Source: developers.openai.com/api/docs/pricing
Google Gemini
| Model | In | Cached in | Out | Notes |
|---|---|---|---|---|
Gemini 4 Carbongemini-4-carbon | $2 | $0.1 | $10 | From 2026-12-01: $4 in, $0.2 cached, $20 out (announced). |
Gemini 4 Argongemini-4-argon | $2 | $0.1 | $10 | From 2026-12-01: $4 in, $0.2 cached, $20 out (announced). |
Gemini 4 Hydrogengemini-4-hydrogen | $1.5 | $0.075 | $7.5 | |
Gemini 3.8 Flashgemini-3.8-flash | $0.75 | $0.075 | $3.75 | default From 2027-01-01: $1.5 in, $0.15 cached, $7.5 out (announced). |
Gemini 3.6 Flashgemini-3.6-flash | $0.75 | $0.075 | $3.75 | From 2027-01-01: $1.5 in, $0.15 cached, $7.5 out (announced). |
Gemini 3.5 Flash-Litegemini-3.5-flash-lite | $0.3 | $0.03 | $2.5 | |
Gemini 3.1 Flash-Litegemini-3.1-flash-lite | $0.25 | $0.025 | $1.5 |
Paid tier, standard. Key: GEMINI_API_KEY. Source: ai.google.dev/gemini-api/docs/pricing
Any OpenAI-compatible endpoint
OpenRouter, Ollama, LM Studio, vLLM, llama.cpp, Groq and others. You pay whatever that endpoint charges; a model on your own computer costs nothing per token.
Set CLIFF_API_KEY and run cliff --provider compatible --base-url <url> --model <name>. Cliff shows token counts; it can only show a price for models in this list.
The same list is at /models.json and /api/models, and inside the agent: cliff models prints it and cliff models --refresh fetches the latest copy. If a price here is wrong, the provider's page wins; tell us on X and we fix the list.