MiniMax M3
MiniMax M3 is the first open-weight model combining three frontiers at once: top-tier coding, a 1M-token context, and native multimodality. Access it through Velokey's unified endpoint with below-list pricing.
Use cases
Deep coding & codebase work · Autonomous agents & tool use
Input
Text, image
Output
Text
Billing
Per token
Start a conversation
Type a message below to begin
Sign up to get $0.5 in free credits — about 20 free images
Pricing details
Transparent usage-based pricing with no hidden fees.
Detailed Pricing
Tiered pricingUsage-based pricing breakdown
MiniMax M3
| Spec | Price | Official/Fal Price(USD) | Savings |
|---|---|---|---|
| Input/1M tokens | 0.24 Credits≈$0.24 | 0.3 Credits | -20% |
| Output/1M tokens | 0.96 Credits≈$0.96 | 1.2 Credits | -20% |
| Cache Read/1M tokens | 0.048 Credits≈$0.048 | 0.06 Credits | -20% |
| Spec | Price | Official/Fal Price(USD) | Savings |
|---|---|---|---|
| Input/1M tokens | 0.48 Credits≈$0.48 | 0.6 Credits | -20% |
| Output/1M tokens | 1.92 Credits≈$1.92 | 2.4 Credits | -20% |
| Cache Read/1M tokens | 0.096 Credits≈$0.096 | 0.12 Credits | -20% |
Billing Rules
- Output price varies by model, resolution, quality, duration, or token usage.
- Uploaded reference files may be charged separately when supported.
- Add-on capabilities such as web search or image search may be billed per request.
- Token-based models are billed by input and output tokens.
- If the primary route is unavailable, Velokey may automatically switch to a stable fallback route when possible.
- Usage is pay-as-you-go with transparent credit deduction.
* Final cost depends on the generated result.
Related models
Explore other models in the same family.
MiniMaxSave 20%MiniMax M2.7
MiniMax M2.7 is a top open-source model for real-world software engineering and agent building, with strong tool skill-adherence and office-document editing.
OpenAISave 20%GPT 5.4 Mini
OpenAI's fast, cost-efficient GPT-5.4 Mini for high-volume production, with vision, long context, and strong tool use at low latency.
AlibabaSave 20%Qwen3.6 Plus
Alibaba's Qwen3.6 Plus, a balanced flagship-tier model with strong reasoning, long context, multimodal input, and agentic tool use.
AnthropicSave 20%Claude Fable 5
Anthropic's Mythos-class frontier model, state-of-the-art across coding, agentic work, vision, and scientific research.
Affordable MiniMax M3 API
Access MiniMax M3 on Velokey — the first open-weight model to combine top-tier coding, a 1M-token context, and native multimodality, with a unified endpoint and below-list pricing for production apps.
Meet MiniMax M3 and its API access
MiniMax M3 is a text model from MiniMax and the first open-weight model to combine three frontier capabilities at once: top-tier coding, a 1M-token context, and native multimodality. It reaches top-tier performance on coding and agentic benchmarks, autonomously decomposing tasks, invoking tools, and reasoning across multiple steps; text and visual semantic spaces are deeply aligned for native multimodal understanding; and its MiniMax Sparse Attention (MSA) architecture supports up to a 1M-token context (512K guaranteed minimum). Velokey provides an affordable, easy-to-integrate MiniMax M3 API: a unified chat endpoint, a clear request flow, and below-list pricing for production systems, agent pipelines, and automation tools.
Three frontiers in one open-weight model
The first open-weight model to combine top-tier coding, a 1M-token context, and native multimodality at once — integrating strengths that used to live in separate models.
View Documentation →
Top-tier coding & autonomous agents
Top-tier performance on coding and agentic benchmarks, with autonomous task decomposition, tool invocation, and multi-step reasoning — it has run a 12-hour ICLR paper reproduction and a CUDA kernel optimization reaching 9.4x speedup.
View Documentation →
1M-token context via Sparse Attention
Built on the MiniMax Sparse Attention (MSA) architecture, it supports up to a 1M-token context with a 512K guaranteed minimum, handling very long codebases and documents efficiently.
View Documentation →
Native multimodality
Deep alignment between textual and visual semantic spaces delivers native multimodal understanding rather than a bolted-on afterthought.
View Documentation →
How to deploy the MiniMax M3 API on Velokey
Get started with just a few steps.
Sign up and get an API Key
Sign in to the Velokey console to generate an API Key that authenticates every request sent to MiniMax M3.
Configure request parameters
Set model to minimax-m3 and submit messages. Use the v2.1 endpoints for coding, agentic tool calling, and very long context.
Integrate and start calling
Send requests to Velokey's unified /v1/chat/completions endpoint to power assistants, agents, and automation, with usage managed in one console.
Real-world use cases for MiniMax M3
From engineering to research, covering many high-autonomy scenarios.
Deep coding & codebase work
Top-tier coding paired with a 1M-token context — fit for development, refactoring, and tuning across an entire codebase.
Autonomous agents & tool use
Autonomous task decomposition, tool invocation, and multi-step reasoning — fit for complex agentic workflows and long-horizon automation.
Research reproduction & engineering
It has run a 12-hour ICLR paper reproduction and a CUDA kernel optimization reaching 9.4x speedup — fit for research and performance engineering.
Very long document & codebase comprehension
A 1M-token context enables retrieval, synthesis, and QA across very long documents and entire codebases.
Multimodal understanding & analysis
Deep text-visual semantic alignment — fit for mixed text-and-image understanding, analysis, and content processing.
Long-horizon automation pipelines
Autonomous task decomposition and tool invocation — fit for automation that needs multi-step execution and tool orchestration.
Why choose Velokey for MiniMax M3
Discount on list price
Call MiniMax M3 at a lower price on Velokey — pay-as-you-go with no forced subscription, reducing upfront cost.
Unified API interface
One API Key accesses MiniMax M3 and other models — no juggling multiple accounts and keys, simplifying integration.
Complete docs and onboarding guidance
Endpoint references, parameter details, and examples help you quickly adopt coding, agentic, and long-context capabilities.
Stable and highly available
Automatic failover and stable fallback routes maintain availability when the primary route is down, keeping production reliable.
API Reference
Complete API call examples and parameter descriptions
Endpoint
https://api.velokey.ai/v1/chat/completionsAuthentication
Bearer YOUR_API_KEYRequest Example
curl https://api.velokey.ai/v1/chat/completions \
-X POST \
-H "Authorization: Bearer YOUR_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "MiniMax-M3",
"messages": [{"role": "user", "content": "Hello!"}],
"temperature": 0.7
}'Response Example
{
"id": "req_abc123",
"model": "MiniMax-M3",
"created": 1234567890,
"data": { ... }
}MiniMax M3 API FAQ
What makes MiniMax M3 unique?
It is the first open-weight model to combine three frontier capabilities at once: top-tier coding, a 1M-token context, and native multimodality — integrating strengths that used to live in separate models.
How good is the long context?
Built on the MiniMax Sparse Attention (MSA) architecture, it supports up to a 1M-token context with a 512K guaranteed minimum, handling very long codebases and documents efficiently.
How does it perform on benchmarks?
It scores 83.5 on BrowseComp (exceeding Opus 4.7's 79.3), ranks #3 overall on PostTrainBench (37.1), and has demonstrated a 12-hour ICLR paper reproduction and a CUDA kernel optimization reaching 9.4x speedup.
How can I access it?
Through the API (v2.1 endpoints), a Token Plan, the MiniMax Code no-code agent interface, with open-source deployment (HuggingFace/GitHub) coming soon.
How is it billed on Velokey?
This page's pricing area shows the current Velokey price, billed by input and output tokens with a discount on list price and no forced subscription.
Start building with MiniMax M3 today
Use one API key to access low-cost, stable AI models through Velokey.