MiniMax Agentic text & coding
MiniMax M2.7
Move from a work request to a dependable tool plan.
MiniMax M2.7 is a MiniMax text route for reasoning, productivity, coding assistance and tool-oriented workflows. Its catalog entry shows a 200K context value and OpenAI-compatible access. Use it for structured project work where the application can validate each step and keep an explicit record of what the model did not know.
- Reasoning
- 200K context
- Tool use
MiniMax M2.7 and the M2.7-highspeed route are separate catalog entries. Do not infer identical quality, pricing or latency; compare both on the same task set if the trade-off matters.
AI Chat
Send a prompt and see the response here.
MiniMax-M2.7Start a conversation with this model and ask follow-up questions.
USD /1M tokens
Pricing
| Group | Input | Output | Cache Read | Cache Write |
|---|---|---|---|---|
| default | $0.315 | $1.26 | $0.063 | $0.394506 |
Public list prices are shown in USD. Final charges may vary by account group and usage tier.
Overview
What is MiniMax M2.7?
The route is cataloged for text input and text output, with reasoning listed as a capability. A long context helps with project material, but a clear request contract and retrieval strategy still determine whether the answer is useful.
For agent work, ask for plans and machine-readable arguments before executing tools. Keep the highspeed variant as an explicit alternative rather than silently switching model IDs.
Start from a failing behavior
Supply the reproduction and relevant files. Ask for a small patch and a test plan that preserves the project's existing conventions.
Turn notes into decisions
Provide dated source notes and an output schema. Keep decisions, assumptions and open questions in separate sections.
Validate before execution
Have the model propose arguments and dependencies first. Let your application reject incomplete or unsafe actions.
From brief to deliverable
Put MiniMax M2.7 to work
Software teams
Trace a bug through a repository
Give the model the error, entry point and nearby modules. Require a causal explanation, a minimal patch and checks that cover the original symptom.
- Tools you need
- Repository access and a test runner if changes will be executed.
- Keep in mind
- Model-generated code is a proposal until the project tests pass.
View a task brief
Inspect the supplied failure and repository context. Explain the likely cause, propose the smallest patch and list a regression check. If execution tools are available, apply only the approved change and report exact command output; otherwise return an unexecuted plan.
Operations teams
Convert operating notes into a runbook
Use a set of internal notes to draft a repeatable runbook with prerequisites, decision points and escalation paths.
- Tools you need
- Document retrieval and access to the team's source of truth.
- Keep in mind
- Do not let the model invent an owner, permission or recovery action that the sources do not specify.
View a task brief
Draft a runbook from the supplied notes. Separate confirmed procedures from recommendations, identify missing owners and add a stop condition whenever a permission or system detail is absent.
Agent and automation builders
Select a tool action from a structured request
Ask the model to classify an incoming request and return a validated tool argument object, with a refusal path for missing fields.
- Tools you need
- An orchestration layer, schema validator and audit log.
- Keep in mind
- The model should not perform or claim a tool action; your application decides whether to execute it.
View a task brief
Given the request and tool schemas, return one validated action proposal or a clarification request. Never invent IDs, permissions or tool results. Use the supplied enum values and include a concise reason code.
Capabilities in context
Key features
| Feature | What you get | Why it matters |
|---|---|---|
| Context capacity | 200K tokens shown in the catalog | Use the available context for project material, but retrieve only what the task needs and preserve source boundaries. |
| Text route | Text input and text output | Design the surrounding workflow for files, images or other modalities when the task needs them. |
| Reasoning | Reasoning is listed by the catalog | Measure accepted results on your own code and document tasks instead of treating a capability label as a quality guarantee. |
| Cache-aware pricing | The route exposes input cache ratios in the catalog | Track repeated prefixes and cache behavior in total-cost comparisons rather than comparing only uncached requests. |
| Highspeed sibling | MiniMax-M2.7-highspeed is a separate route | Pin the exact model ID and benchmark latency, correction rate and cost before changing traffic. |
Model identity, modalities and route metadata come from TokenHot's catalog. Compare the standard and highspeed IDs explicitly; capability labels do not replace an evaluation.
Audience & fit
Where MiniMax M2.7 fits
Engineering teams needing a practical coding partner
Use it for repository questions, test planning and small patches where a human reviews the result before merge.
Operations teams building runbooks
Ground the request in dated notes and keep missing ownership or permissions visible.
Agent builders
Use a schema-first tool contract and a refusal path. The application remains responsible for authorization and execution.
Cost context
Plan the cost of a useful result
Include validation cost
A cheaper response that needs repeated repair may cost more than a slower first pass. Count retries, schema failures and human correction.
Watch cache behavior
Large shared instructions can make cache settings material. Measure repeated-prefix workloads separately from one-off prompts.
Review the full group table
TokenHot account groups can change the effective rate. Use the dynamic pricing table for the group attached to your request.
Use the current pricing table for available rates and account groups. Model IDs and prices come from TokenHot’s catalog; provider pricing and subscriptions are separate.
API
Code examples
curl --location --request POST 'https://api.tokenhot.ai/v1/chat/completions' \
--header 'Authorization: Bearer <token>' \
--header 'Content-Type: application/json' \
--data-raw '{
"model": "MiniMax-M2.7",
"messages": [
{
"role": "user",
"content": "What model are you"
}
]
}'FAQ
Frequently asked questions
What is MiniMax M2.7 best suited for?
MiniMax M2.7 is suited to the input, output, and capability types listed on this page. Test production workloads before choosing it for a critical system.
How is MiniMax M2.7 priced?
Pricing is shown in the pricing section above. Actual cost depends on usage volume, account group, and the request payload.
How can I call MiniMax M2.7?
Use the model ID shown above with one of the supported API protocols. Code examples are provided when usage data is available.
Is M2.7-highspeed the same model?
It is a separate TokenHot route. Compare the two model IDs on the same prompts before making a speed or quality claim.
Can M2.7 execute tools by itself?
No. Tool use requires your application to expose schemas, authorize calls and return results. The model can propose arguments and next steps.
Is the 200K context a guaranteed request size?
It is the catalog context value. Effective input, output and endpoint limits can be lower, so verify the request budget in your client.
What should I test first?
Start with a small set of coding, document and tool-routing tasks where acceptance criteria are explicit and easy to replay.
Get started
Build with MiniMax M2.7
API