Frontier coding
Strong SWE-Bench Pro and Terminal-Bench results for real coding and debugging.
M-Chat: Chat with MiniMax M3 Online
Use M-Chat, an independent web app for MiniMax M3-powered coding, agent workflows, long-context research, and 1M-token prompts.
M-Chat is independent and not affiliated with MiniMax. No setup required. Thinking mode and web search available.
Use MiniMax M3, enable web search or Thinking when needed, then start with a real prompt.
Starter prompts
Overview
How M-Chat Uses MiniMax M3 AI
M-Chat is an independent interface built around MiniMax M3, an open-weight frontier model for coding, agentic tool use, and long-context work. The model accepts native text, image, and video input, and its MSA architecture cuts per-token compute at 1M context to about 1/20 of the previous generation.
Open-weight frontier model
Released with open weights, focused on coding and agentic tasks.
MSA architecture, 1M context
Multi-Sparse-Attention keeps 1M-token context affordable — about 1/20 the per-token compute of the prior generation.
Native multimodal input
Accepts text, image, and video input with text output.
Why M-Chat
Why Use M-Chat for MiniMax M3
Why M-Chat makes MiniMax M3 practical for day-to-day engineering and agent work.
Low cost
$0.30 / 1M input and $1.20 / 1M output — frontier quality at a fraction of typical cost.
1M-token context
Hold whole repos, long documents, and multi-step agent traces in one context.
Agentic tool use
Built for multi-step tool calling, MCP, and autonomous workflows.
Open weights
Open-weight release you can self-host and audit.
Fast long-context inference
MSA architecture keeps 1M-context responses fast and affordable.
M-Chat Features
The core capabilities M-Chat adds around MiniMax M3 access.
Thinking mode
Toggle extended reasoning for harder coding, planning, and analysis.
Web search
Server-side web search injects current evidence before the model answers.
Multimodal input
Send text, images, and video as described by the model's capabilities.
Coding and agents
Write, refactor, and debug code, or drive multi-step agent tasks.
1M context
Work across large codebases and long documents without losing context.
API access
Use MiniMax M3 in your own apps.
MiniMax M3 Benchmark Table
Full MiniMax M3 benchmark table from the official release, covering coding, cowork agent, GUI, multimodal, and reasoning evaluations.
| Benchmark | MiniMax M3 | MiniMax M2.7 | Claude Opus 4.7 | GPT 5.5 | Gemini 3.1 Pro | Claude Sonnet 4.6 | DeepSeek V4 Pro | GLM 5.1 Thinking | Kimi K2.6 Thinking |
|---|---|---|---|---|---|---|---|---|---|
| Coding | |||||||||
| SWE-Bench Verified | 80.5 | 79.9 | 87.6 | 82.9 | 80.6 | 79.6 | 80.6 | - | 80.2 |
| SWE-Bench Pro | 59.0 | 56.2 | 64.3 | 58.6 | 54.2 | - | 55.4 | 58.4 | 58.6 |
| Terminal Bench 2.1 | 66.0 | 51.1 | 66.1 | 78.2 | 70.3 | - | - | - | - |
| SWE Atlas-QnA | 37.9 | 11.29 | 45.16 | 45.43 | 13.5 | 31.20 | - | - | - |
| nl2repo | 42.13 | 34.99 | 56.28 | 52.9 | 21.62 | - | 35.5 | 41 | 42.8 |
| SWE Atlas-Test Writing | 30.83 | 18.89 | 38.21 | 42.59 | 29.84 | 31.76 | - | - | - |
| SWE-fficiency | 34.8 | 13.98 | 42.2 | 46.6 | 19.7 | - | - | - | - |
| LiveSQLBench | 40.17 | 33.17 | 41.00 | 40.17 | 39.83 | - | - | - | - |
| CL-bench | 20.48 | 15.38 | 22.92 | 25.38 | 21.06 | - | - | - | - |
| VIBE-V2 | 50.12 | 37.89 | 55.87 | 50.50 | 28.00 | - | - | - | - |
| SVG-Bench | 63.7 | 48.0 | 62.3 | 58.2 | 59.2 | - | - | - | - |
| PostTrainBench | 37.1 | 13.1 | 42.4 | 39.3 | 15.2 | - | - | - | - |
| KernelBench Hard | 28.8 | 10.5 | 30.7 | 20.9 | 18.6 | - | - | - | - |
| PaperBench | 52.6 | 30.6 | 58.5 | 57.5 | 46.7 | - | - | - | - |
| Cowork (Agent) | |||||||||
| BrowseComp | 83.52 | 76.3 | 79.3 | 84.4 | 85.9 | 74.7 | 83.4 | 79.3 | 83.2 |
| DRACO | 73.23 | 66.77 | 77.7 | - | - | 75.8 | - | - | - |
| GDPval rubrics | 74.78 | 66.44 | 79.8 | 80.66 | 57.82 | 75.65 | 70.32 | 68.26 | 65.12 |
| BankerToolBench | 76.12 | 63.89 | 81.34 | 70.04 | 67.03 | - | - | - | - |
| OfficeQA Pro | 45.1 | - | 43.6 | 52.6 | 18.1 | - | - | - | - |
| SpreadSheetBench-v1 | 89.35 | 84.92 | 88.49 | 88.11 | 56.06 | - | 84.9 | 85.2 | 84.5 |
| YC-Bench | 2.10M | 0 | 2.19M | 1.28M | 1.05M | - | - | - | - |
| LOCA-Bench (256k) | 49.3 | 0 | 57 | - | - | - | - | - | - |
| MCP Atlas | 74.2 | 49.4 | 77 | 75.3 | 69.2 | 61.3 | 73.6 | 71.8 | 66.6 |
| Apex-Agents | 27.7 | 5.6 | 37.2 | 41.7 | 33.4 | 26.2 | - | - | - |
| Claw-Eval | 74.5 | 49.7 | 71.6 | - | 57.8 | 68.3 | 58.4 | 62.7 | 61.5 |
| GUI | |||||||||
| OSWorld-Verified | 70.06 | - | 82.8 | 78.7 | 76.2 | 72.5 | 80.6 | - | 73.1 |
| MultiModal | |||||||||
| OmniDocBench | 91.6 | - | 89.3 | 87.5 | 88.1 | 86.9 | - | - | - |
| MMMU-Pro | 78.1 | - | 77 | 81.2 | 80.5 | 74.5 | - | - | 79.4 |
| Video-MMMU | 84.6 | - | 83 | 86.4 | 87.9 | - | - | - | - |
| VideoMME (w/ sub) | 85.4 | - | - | 89.4 | 87.9 | - | - | - | - |
| Reasoning | |||||||||
| IMO 2025 | 35 / 42 | - | - | - | - | - | - | - | - |
| USAMO 2026 | 36 / 42 | - | 52.8% | 98.21% | 74.40% | - | - | - | - |
Source: MiniMax M3 official release full benchmark table. Dashes match empty cells in the source table.
Updated 2026-06-01Pricing
M-Chat Pricing
Choose an M-Chat plan for hosted MiniMax M3 access.
Free
$0 / month
Try M-Chat before paying. Best for a quick model check, not regular work.
Good for quick testing before choosing a paid plan.Includes
- 10 credits after sign-in
- M-Chat text access to MiniMax M3
- Thinking and web search available
- Saved chat history
- Upgrade when you need more credits
Lite
$5.99 / month
Affordable monthly M-Chat credits for light but recurring usage.
Entry paid plan with predictable monthly credits.Includes
- 300 credits per month
- Best for occasional weekly chat
- MiniMax M3: 3 credits per message
- Thinking: +1 credit
- Web search: +2 credits
- Email support
- No ads
Recommended
Starter
$12.99 / month
Balanced monthly plan for daily research, coding, and web-search-assisted chat.
Recommended monthly plan for regular M-Chat users.Includes
- 900 credits per month
- 3x Lite monthly credits
- MiniMax M3: 3 credits per message
- Thinking: +1 credit
- Web search: +2 credits
- Priority response speed
- Saved chat history
- Email support
Pro
$29.99 / month
Higher monthly M-Chat credit pool for heavier chat and long-context work.
For teams, power users, and longer research sessions.Includes
- 2400 credits per month
- 8x Lite monthly credits
- Best monthly unit value
- MiniMax M3: 3 credits per message
- Thinking: +1 credit
- Web search: +2 credits
- Priority support
- No ads
- Longer research sessions
- Built for heavier daily usage
M-Chat FAQ
Common questions about M-Chat, MiniMax M3, Thinking, web search, and local validation.
What is M-Chat?
M-Chat is an independent web chat and research site for MiniMax M3. It is not MiniMax official infrastructure.
Which model does the chat use?
The chat uses MiniMax M3.
Does MiniMax M3 support multimodal input?
MiniMax M3 is described as supporting text, image, and video inputs with text output. M-Chat is validated for text chat.
What does Thinking do?
Thinking enables extended reasoning for prompts that need deeper analysis.
How does web search work?
The server runs a web search and provides the results as context to MiniMax M3 before the answer is generated.
Is MiniMax M3 free to use?
MiniMax M3 is released as an open-weight model, and M-Chat gives every signed-in user 10 free credits to try it. Beyond that, access is paid — M-Chat sells credit plans, and using the model through other providers is billed by them per token.
How much does MiniMax M3 cost?
Third-party providers have listed MiniMax M3 around $0.30 per 1M input tokens and $1.20 per 1M output tokens during its launch discount. The official MiniMax API is tiered by input length, with higher rates above 512K tokens. Verify live pricing before budgeting.
Is MiniMax M3 open source?
MiniMax M3 ships as an open-weight model you can download and self-host, but the release stops short of a full open-source license. Check MiniMax's terms before building a commercial product on the weights.
Is MiniMax M3 good for coding?
MiniMax M3 targets coding and agentic work, scoring 59.0% on SWE-Bench Pro and 66.0% on Terminal-Bench 2.1 in MiniMax's release. It narrowly leads the open-access group on SWE-Bench Pro, though closed models like Claude Opus 4.8 score higher.
How does MiniMax M3 compare to GPT-5.5 and Claude Opus 4.8?
MiniMax reports M3 just ahead of GPT-5.5 on SWE-Bench Pro and stronger on BrowseComp. Claude Opus 4.8 leads on raw coding (about 69.2% on SWE-Bench Pro) but costs far more. M3's pitch is near-frontier quality at a fraction of the price.
What is MiniMax M3's context window?
MiniMax M3 has a 1M-token context window, larger than many peers such as Kimi K2.6 at 256K. That makes it practical for full-repository analysis, long transcripts, and tool-heavy agent sessions inside a single prompt.
How do I use the MiniMax M3 API?
MiniMax M3 is available through MiniMax's official API and as an open-weight model you can self-host. M-Chat is a hosted chat interface, not an API to build on — to integrate M3 into your own app, use MiniMax's API or the open weights.
Latest MiniMax M3 Guides
Report notes, benchmark, pricing, and model comparison guides for MiniMax M3.
MiniMax M3 Benchmark: Coding, Agentic, and Long-Context Scores
MiniMax M3 benchmark results on SWE-Bench Pro, Terminal-Bench, BrowseComp, and MCP Atlas, how the model compares to GPT-5.5 and Claude Opus 4.8, and what early developers are saying.
Jun 1, 2026
MM-Chat Team
MiniMax M3 Price: API Token Costs and How to Budget
MiniMax M3 price across OpenRouter and the official MiniMax API, how it compares to Claude Opus 4.8 and GPT-5.5 on cost, M-Chat credits, and what the community says.
Jun 1, 2026
MM-Chat Team
MiniMax M3 vs DeepSeek V4 Pro: Benchmarks and Practical Choice
MiniMax M3 vs DeepSeek V4 Pro: M3 leads SWE-Bench Pro 59.0% vs 55.4%, DeepSeek leads Terminal-Bench. Specs, routing, and what to test before production.
Jun 1, 2026
MM-Chat Team
MiniMax M3 vs GLM 5.1: Coding and Agentic Comparison
MiniMax M3 vs GLM 5.1: M3 leads every reported row, including SWE-Bench Pro 59.0% vs 58.4%. Specs table, when to choose each, routing, and FAQ.
Jun 1, 2026
MM-Chat Team
MiniMax M3 vs Kimi K2.6: Long-Context and Coding Comparison
MiniMax M3 vs Kimi K2.6: 1M vs 256K context and SWE-Bench Pro 59.0% vs 58.6%. Specs, when to choose each, routing, and what to test before production.
Jun 1, 2026
MM-Chat Team