MiniMax M2.7 API: pricing and free start

Prices per 1 million tokens as of November 2, 2026:

The difference from the cheapest provider of the same model is roughly ×34 on input and ×45 on output. JoinGonka's price on the page is live: it comes from the gateway API and changes with the GNK rate.

MiniMax is moving subscriptions to M3.1 Flash and offering M3 to new users — those are different models, not available on the Gonka network.

1. Key. Sign up at JoinGonka Gateway and create a jg-… key in your dashboard. New accounts get 3M free tokens.

2. Request. The endpoint is https://gate.joingonka.ai/v1, the model is MiniMaxAI/MiniMax-M2.7. If your code already worked with the MiniMax API through the OpenAI SDK, all you change is the endpoint, the key and the model name:

from openai import OpenAI

client = OpenAI(api_key="jg-your-key", base_url="https://gate.joingonka.ai/v1")
stream = client.chat.completions.create(
    model="MiniMaxAI/MiniMax-M2.7",
    messages=[{"role": "user", "content": "Summarize this support chat in three sentences: ..."}],
    stream=True,
)
for chunk in stream:
    if chunk.choices and chunk.choices[0].delta.content:
        print(chunk.choices[0].delta.content, end="", flush=True)

For clients on the Anthropic Messages API — including code written for https://api.minimax.io/anthropic — the endpoint is https://gate.joingonka.ai, without /v1.

3. Coding tools. npx @joingonka/setup --model minimax writes the gateway and model into the config of Claude Code, Codex CLI, OpenCode and other tools, and prints values to paste for Cursor, Cline and Aider.

MiniMax M2.7 is an MoE model from the Shanghai lab MiniMax. The hallmark of the M series is efficient linear attention, which makes long context cheaper. On the Gonka network the model has been running since May 2026, and its working parameters are set by the deployment on the hosts:

• Context — 180,000 tokens, input and answer combined: that's how the network's hosts are configured. At MiniMax and OpenRouter providers the window is around 200,000.
• Answer — up to 8,192 tokens, verified with long generations right up to the ceiling.
• Text only: the network doesn't accept images yet.

Current limits are returned by GET https://gate.joingonka.ai/v1/models, availability — by the gateway status page.

M2.7 is still handy for short, fast answers, but its generation is on the way out: MiniMax has already replaced it with new models, and the model lineup on the Gonka network is changing too. For a new project it makes more sense to start with GLM-5.3 Flash or DeepSeek V4 Flash — the price in JoinGonka is the same, and switching is one model line.

• Code written for M2.7 at MiniMax. It moves to JoinGonka by changing the endpoint, key and model name, with no need to rework prompts for a new model — as long as M2.7 is available on the Gonka network.
• Short answers and bulk requests: classification, fact extraction, drafts.
• Transcripts and logs: lots on the input, a short summary on the output, so the 8,192-token answer ceiling isn't a problem.

When you need a different model: long answers — DeepSeek V4 Flash, up to 32,768 tokens; reasoning tasks — GLM-5.3 Flash. The network's lineup changes by vote of its participants, and the current list is returned by GET https://gate.joingonka.ai/v1/models.