Back to models

MiniMax M3

Available Serverless
MiniMax
Modalities
Input / Output Price
$0.30/$1.20/ 1M
Context Length
512K tokens
Parameters
-

Providers

Different companies host the same model. DOS.AI routes your request to one of them based on the routing mode you pick — Balanced (price + speed), Nitro (fastest), or Exacto (highest tool-calling accuracy).

Reasoning effortAll
ProviderInput /1MOutput /1MContext
MiniMax
$0.30$1.20512K

Pricing

Transparent pass-through token pricing per million tokens. DOS.AI strictly enforces a Zero-Markup retail policy.

Input / Prompt
$0.30/ 1M tokens
Published official list price
Output / Completion
$1.20/ 1M tokens
Zero-markup retail rate
Prompt Cache Read
$0.03/ 1M tokens
Up to 90% cache discount
Context Window
512Ktokens
Full context window supported

Uptime & Reliability

Live upstream health and gateway reliability. Automatic multi-provider failover protects your production traffic.

All Endpoints Operational
SLA Target: 99.99%Last 30 Days: 100.0%
30 days ago100.0% uptime over past monthToday

Benchmarks & Evaluations

Standardized evaluations & blind battle ratings. Higher is better — rank percentile reflects position among frontier models on DOS.AI.

37.4
Top 22%
Artificial Analysis Intelligence Index
Better than 78% of models compared
54.6
Top 22%
Artificial Analysis Coding Index
Better than 78% of models compared
42.1
Top 22%
Artificial Analysis Agentic Index
Better than 78% of models compared

Reasoning

Complex multi-step logic, graduate-level science, and frontier reasoning

3 benchmarks
GPQA Diamond
61.9%
Graduate-level scientific reasoning
HLE
29.6%
Humanity's Last Exam frontier benchmark
AA-LCR
54.5%
Long context reasoning evaluation

Coding

Autonomous repository bug fixing, algorithms, and scientific computation

3 benchmarks
SciCode
40.6%
Python programming for scientific computing
SWE-bench Verified
37.4%
Autonomous multi-file bug fixing and test passage
HumanEval
64%
Algorithmic code generation pass@1

Knowledge & Reliability

Factual hallucination minimization, tool calling, and domain accuracy

4 benchmarks
AA-Omniscience Accuracy
45.2%
Proportion of correctly answered questions
AA-Omniscience Non-Hallucination Rate
28.1%
Rate of avoiding hallucination among non-correct responses
MMLU-Pro
64%
Multi-discipline reasoning & higher education knowledge
Exacto BFCL Tool Calling
73.7%
Strict schema compliance and function execution
Standardized evaluations sourced from Artificial Analysis and open frontier model research benchmarks.

API Integration

Use the DOS API to integrate MiniMax M3 into your applications. Compatible with standard OpenAI SDKs for easy drop-in migration.

</>Run inference
curl https://api.dos.ai/v1/chat/completions \
  -H "Authorization: Bearer $DOS_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "minimax-m3",
    "messages": [
      {
        "role": "user",
        "content": "Hello!"
      }
    ]
  }'