minimax-m2MiniMax-M2 is a compact, high-efficiency model with 10B active (230B total) parameters, optimized for coding and agentic workflows. It delivers near-frontier reasoning and tool use, excels at multi-file coding tasks and compile-run-fix loops, and performs strongly on benchmarks like SWE-Bench and Terminal-Bench. It also handles long-horizon planning and recovery in agent evaluations, ranking among the top open models across reasoning domains. With fast inference and low cost, it’s ideal for large-scale agents and developer assistants — and works best when reasoning is preserved across turns.
Select an endpoint and copy a working example for this model.
from openai import OpenAI client = OpenAI( api_key="YOUR_API_KEY", base_url="https://api.apertis.ai/v1") response = client.chat.completions.create( model="minimax-m2", messages=[ {"role": "user", "content": "Hello!"} ], max_tokens=1024, temperature=0.7) print(response.choices[0].message.content) # Optional: Enable context compression to reduce token usage# response = client.chat.completions.create(# model="minimax-m2",# messages=[{"role": "user", "content": "Hello!"}],# extra_body={"compression": {"enabled": True, "model": "gpt-4.1-mini"}}# )modelmessagesmax_tokenstemperaturetop_pstreamtoolsreasoning_effortstream_optionsthinkingextra_bodyUse these namespaced identifiers in Cursor IDE to avoid conflicts with built-in models.
See how this model compares to others from the same provider.
MiniMax-M3 is a multimodal foundation model from MiniMax, supporting text, image, and video inputs with text output and a 1M-token context window. It is designed for long-horizon agentic workflows, coding, and tool-driven task execution, enabling sustained reasoning across complex tasks. Built on MiniMax Sparse Attention (MSA), the model dramatically improves long-context efficiency by replacing full attention with KV-block selection, reducing compute costs at 1M-token contexts while maintaining strong performance. Trained as a native multimodal model and optimized for multi-turn, production-style collaboration, MiniMax-M3 excels at extended, multi-step workflows rather than single-turn interactions.
MiniMax-M2.7 is a next-generation large language model designed for autonomous, real-world productivity and continuous improvement. It incorporates advanced multi-agent collaboration, enabling the model to plan, execute, and iteratively refine complex tasks across dynamic environments. Built for production-grade workflows, M2.7 supports tasks such as live debugging, root cause analysis, financial modeling, and full document generation across Word, Excel, and PowerPoint. With strong benchmark performance—including 56.2% on SWE-Pro, 57.0% on Terminal Bench 2, and 1495 ELO on GDPval-AA—it sets a new standard for multi-agent systems in real-world digital workflows.
MiniMax-M2.5-Lightning is the high-speed variant of the M2.5 series, optimized for low latency, real-time responsiveness, and high-frequency workloads. It retains the core planning and execution strengths of M2.5 while further improving inference efficiency and response speed, making it ideal for interactive applications, rapid coding assistance, and workflow automation. With enhanced cost efficiency and reduced latency, M2.5-Lightning is particularly well suited for high-throughput, always-on deployments and production environments where speed and scalability are critical.
See how this model compares to others from the same provider.
MiniMax-M3 is a multimodal foundation model from MiniMax, supporting text, image, and video inputs with text output and a 1M-token context window. It is designed for long-horizon agentic workflows, coding, and tool-driven task execution, enabling sustained reasoning across complex tasks. Built on MiniMax Sparse Attention (MSA), the model dramatically improves long-context efficiency by replacing full attention with KV-block selection, reducing compute costs at 1M-token contexts while maintaining strong performance. Trained as a native multimodal model and optimized for multi-turn, production-style collaboration, MiniMax-M3 excels at extended, multi-step workflows rather than single-turn interactions.
MiniMax-M2.5 is a state-of-the-art large language model designed for real-world productivity and digital work environments. Building on the coding strengths of M2.1, it expands into general office workflows, demonstrating strong capability in generating and operating Word, Excel, and PowerPoint files, switching context across software environments, and collaborating effectively with both human users and agent systems. Trained on diverse real-world working scenarios, M2.5 combines strong planning ability with improved token efficiency, enabling more effective task execution. With strong benchmark performance—including 80.2% on SWE-Bench Verified, 51.3% on Multi-SWE-Bench, and 76.3% on BrowseComp—it is well suited for productivity automation, coding workflows, and agent-driven knowledge work.
MiniMax-M2.5-Lightning is the high-speed variant of the M2.5 series, optimized for low latency, real-time responsiveness, and high-frequency workloads. It retains the core planning and execution strengths of M2.5 while further improving inference efficiency and response speed, making it ideal for interactive applications, rapid coding assistance, and workflow automation. With enhanced cost efficiency and reduced latency, M2.5-Lightning is particularly well suited for high-throughput, always-on deployments and production environments where speed and scalability are critical.
MiniMax-M2.1 is a lightweight, state-of-the-art model optimized for coding and agentic workflows, using just 10B activated parameters to deliver strong real-world performance with low latency and high cost efficiency. It improves on M2 with cleaner outputs and faster responses, leads in multilingual coding benchmarks (49.4% Multi-SWE-Bench, 72.5% SWE-Bench Multilingual), and serves as a versatile agent core for IDEs, coding tools, and general applications.
Initialized observational baseline with no recorded failures