MiniMax-M2.7 API pricing in 2026: what 1M tokens costs

MiniMax-M2.7 costs $0.30 per 1M input tokens and $1.20 per 1M output tokens on OpenRouter (reference price, 26 October 2026). On CheapTokens the same model is cheaper than OpenRouter in total, with input and output counted together: $0.05 in and $0.15 out per 1M tokens right now, and the table below recalculates on its own if the prices change.

That's the short answer. The useful part is what it means for your bill, because a request is paid for twice: once for what you send, once for what the model writes back.

What MiniMax-M2.7 costs per 1M tokens#

A token is a few characters of text; the bill counts them separately for input (your prompt and history) and output (the answer). If that split is new to you, input vs output tokens and what a token is take five minutes.

The prices you can check:

• OpenRouter, reference price as of 26 October 2026: $0.30 input, $1.20 output per 1M tokens.
• CheapTokens: the live prices in the first paragraph and in the examples below.

On CheapTokens the model is MiniMaxAI/MiniMax-M2.7 with a 180K context window and up to 8,192 output tokens per answer. The price is fixed per model, with no peak-hour surcharge, and one balance covers every model in the catalog.

Example 1: a support bot, 1,000 requests#

Say your bot gets 1,000 tickets a day. Each request carries about 300 input tokens (the question plus a short instruction) and gets about 150 output tokens back.

• Input: 1,000 × 300 = 300,000 tokens
• Output: 1,000 × 150 = 150,000 tokens

On OpenRouter that is 0.3 × $0.30 + 0.15 × $1.20 = $0.09 + $0.18 = $0.27 (reference prices of 26 October 2026). Here is the same day on live prices:

Notice where the money goes: the answer is half as long as the question, yet output is the larger line, because each output token is priced higher than an input token.

Example 2: an agent loop, 200 steps#

Agents in a coding tool behave differently. Every step resends the files, the tool results and the history, so input is far bigger than output. Take 200 steps, each with about 20,000 input tokens and 500 output tokens.

• Input: 200 × 20,000 = 4,000,000 tokens
• Output: 200 × 500 = 100,000 tokens

OpenRouter: 4 × $0.30 + 0.1 × $1.20 = $1.20 + $0.12 = $1.32 (reference prices of 26 October 2026). On CheapTokens:

A big input is the normal state of an agent, not a mistake; the Aider cost breakdown shows what such a tool sends on every step.

Send the first request#

The API is OpenAI- and Anthropic-compatible. Use the full model id: short names from price tables are not accepted by the API, and ids differ between providers.

curl https://api.cheaptokens.dev/v1/chat/completions \
  -H "Authorization: Bearer $CHEAPTOKENS_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "MiniMaxAI/MiniMax-M2.7",
    "max_tokens": 300,
    "messages": [{"role": "user", "content": "Summarise in one sentence what a token is."}]
  }'

The response has a usage block with prompt_tokens and completion_tokens. Multiply those by the prices in the table and you have the exact cost of the call. A finish_reason of length means the answer hit max_tokens; see why an answer stops mid-sentence.

Setup details are in the docs, the full price list is on the pricing page, and the model page is MiniMax on CheapTokens.

Which model for which job#

For a multi-step agent with a big input, the 180K context and the agent-run price in the chart above are what matter. For a long report in one answer, the limit is 8,192 output tokens per answer, so split bigger jobs into parts. For a side-by-side with other models, read GLM vs DeepSeek vs MiniMax, or the pricing guides for DeepSeek and GLM.