Kyma APIKyma API
MiniMax

MiniMax

MiniMax M3

MiniMax M3 is a frontier model optimized for agentic coding workflows and long-context multimodal analysis. Developers building autonomous agents or debugging complex repositories should use it when they need reliable tool execution across a million-token window.

Modalities

Text+Image → Text

Below release

fp4 · fp8

Weights

Published

MiniMaxAI/MiniMax-M3

Where it sits in the catalogue Kyma measures

Every number against every text model Kyma prices per token — a stated rule, not a chosen line-up.

Price · input#22/71

$0.405

output $1.62 · cached $0.0756

median $1.013 · best $0.0494

Price · output#22/70

$1.62

per 1M tokens generated

median $3.531 · best $0.218

Availability · 30d#32/54

99.62%

264 observations since 2026-06-10

median 99.9% · best 100.0%

Throughput#30/65

38 tok/s

probe median, one fixed prompt

median 35.7 tok/s · best 365.1 tok/s

Response time#25/65

2.06 s

probe median, to a complete answer

median 2.96 s · best 0.89 s

Context

1.05M

max output 32,768

published by MiniMax, not measured here

Tick above each rail is this model, below it the other 70. Dashed rule is the field median, solid is its best. Better is left; the axis stops at the 90th percentile, so a few models sit past its right edge.

Usage

How much this model is actually called here.

Rank

#23

of 101 active models

Tokens served

27.0M

all-time

Platform share

0.6%

of all tokens

Tokens · last 14 daysSep 26 → Oct 9

Two clocks, and why they disagree

Kyma measures this model twice. Both are real; they answer different questions.

Probe · every 6h · 30 days

2.06s to answer

One fixed prompt, on a schedule, to every model. Comparable, because the model is the only thing that changes.

Observations
264
Answered by a substitute
1

Real traffic · last 7 days

2.99s to answer

Your requests, at the lengths clients actually send. Not comparable between models, but it is what running this one feels like.

Requests
38
Completed
100%
p95
18.1 s

Caching, as realised · last 7 days

12.3% of input cached

The share of input that actually hit cache, so the effective rate below is what was charged, not a best case.

Cached input
4.6K
Fresh input
32.9K
List input
$0.405 /1M
Effective input
$0.36443 /1M

The gap is prompt length, not the model degrading. Use the probe figure to choose between models, the traffic figure to budget for your own.

Pricing

Pay per token. Cached input bills at this model’s own cached rate, listed below.

$0.405 /1M input$1.62 /1M output
Hobby10 req/day · 2K in / 500 out
~$0.49/mo
Production1,000 req/day · 2K in / 500 out
~$48.60/mo
Scale20,000 req/day · 2K in / 500 out
~$972/mo
+ Estimate your workload
1,000
2,000
500
30%

Estimated monthly cost

$42.67

$1.42 / day on MiniMax M3

Same workload on:

MiniMax M2.5$48.60+14%
MiniMax M2.7$48.60+14%

Estimates use list pricing with cached input at this model's own cached rate. Actual bills depend on real token counts, and every response includes its exact cost.

When to use MiniMax M3

Updated 2026-07-29

Where this model earns its cost — and where it doesn't.

MiniMax M3 uses MSA sparse attention to process text, image, and video inputs within a 1,048,576-token context window. It supports tool calling, structured outputs, and explicit reasoning traces, making it suitable for multi-step coding tasks and repository-level analysis.

On Kyma, the model runs through an OpenAI-compatible endpoint with automatic request failover and exact cost reporting in the usage.cost field. Prompt caching applies to repeated prefixes.

Output generation caps at 32,768 tokens, and generation is slow. It is not designed for high-speed conversational streaming or low-latency real-time applications.

Agentic Code Generation

Handles multi-step repository edits and tool chaining across long context windows.

Multimodal Input Analysis

Processes text, image, and video inputs to extract structured data or debug visual workflows.

Long-Horizon Debugging

Maintains state over extended sessions to trace errors across large codebases.

Structured Output Parsing

Returns deterministic JSON or XML formats for downstream pipeline integration.

Not ideal for: Avoid this model for low-latency chat or high-throughput streaming where sub-second response times are required.

How it compares

Against the peers people actually weigh it against.

SpecMiniMax M3MiniMax M2.5MiniMax M2.7
Input /1M$0.405$0.405$0.405
Output /1M$1.62$1.62$1.62
Context1M197K205K
ToolsYesYesYes
ReasoningYesYesYes
Throughput38 tok/s48 tok/s51 tok/s

Quick start

Up and running in under two minutes.

  1. 1

    Create an API key

    Sign up and grab a key from the dashboard — $0.50 free credit on the free tier, which covers MiniMax M3. No card required.

    Get API key →
  2. 2

    Make your first request

    Drop in your key and send a chat completion — fully OpenAI-compatible.

    curl https://kymaapi.com/v1/chat/completions \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "minimax-m3",
    "messages": [
      {"role": "user", "content": "Explain prompt caching in one paragraph."}
    ]
  }'
  3. 3

    Stream responses

    Add "stream": true to receive tokens as they arrive.

    curl https://kymaapi.com/v1/chat/completions \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "minimax-m3",
    "stream": true,
    "messages": [{"role": "user", "content": "Hello!"}]
  }'

FAQ

Common questions about this model.

What is the context window of MiniMax M3?

MiniMax M3 has a 1M-token context window — roughly 1542 pages of text in a single request.

How much does the MiniMax M3 API cost?

$0.405 per 1M input tokens and $1.62 per 1M output tokens, with cached input at $0.0756/1M — 81% off the input rate on repeated prompt prefixes. No subscription; you pay only for what you use.

Does MiniMax M3 support function calling?

Yes — MiniMax M3 supports tool/function calling and structured outputs (JSON mode), so it works with agent frameworks out of the box.

Are the weights for MiniMax M3 publicly available?

Yes. MiniMax publishes MiniMax M3's weights as MiniMaxAI/MiniMax-M3, so you can download and run the model yourself (https://www.minimax.io/blog/minimax-m3, read 2026-08-14). Kyma serves it because it is convenient and has failover behind it, not because it is the only way to reach it.

Is MiniMax M3 ever served below the precision its creator released it at?

Sometimes. MiniMax M3 was released by MiniMax at bf16 (https://huggingface.co/MiniMaxAI/MiniMax-M3/blob/main/config.json, read 2026-08-14), and at least one route serving it here reports fp4 or fp8 — narrower than that. Weights compressed below the release usually answer close to it, but it is not the identical artefact, and nothing in the response tells you which one answered. Kyma checked the routes on 2026-09-21, so it is your call rather than a silent one.

How do I use MiniMax M3?

Kyma is OpenAI-compatible: point your SDK's base URL at https://kymaapi.com/v1, use your Kyma API key, and set the model to minimax-m3. Signing up is free and includes $0.50 of credit on the free tier, which covers this model — no card required.

Does it support image and video inputs?

Yes, the model accepts text, image, and video inputs but outputs text only.

How does prompt caching work on Kyma?

Repeated prompt prefixes are billed at this model's cached input rate, reducing costs for iterative agent loops.

Can I rely on it for structured JSON outputs?

Yes, it natively supports structured outputs and tool calling, which you can enforce via system prompts or API parameters.

Start with $0.50 free credit on the free tier — no card required.Create account →

More models by MiniMax

ModelContextInputOutput
MiniMaxMiniMax M2.7205K$0.405$1.62
MiniMaxMiniMax M2.5197K$0.405$1.62
MiniMaxMiniMax Speech Turbo—$0.04 / 1K char
MiniMaxMiniMax Speech HD—$0.07 / 1K char
MiniMaxMiniMax Image 01—$0.005 / image