MiniMax M2.5 — EU Hosted
Sonnet-level performance. 10x lower cost.
The #2 model on SWE-bench Verified, now running on EU infrastructure in Germany. 160K context, 400+ tokens/sec, built-in reasoning. Your data never leaves Europe.
More Than Just Coding
While M2.5 dominates coding benchmarks (#2 on SWE-bench Verified), it was trained with senior professionals in finance, law, and social sciences. 160K context window. Native chain-of-thought reasoning. Designed for the full spectrum of agentic workflows.
Financial Analysis
End-to-end research, data fetching, formatted reports
Web Research
Deep exploration across information-dense pages
Document Processing
Legal review, contract analysis, compliance checking
Domain-Specific Agents
Combine Skills with industry SOPs for automation
Agentic Coding
#2 on SWE-bench Verified
The industry standard for evaluating AI coding assistants on real GitHub issues. Works out of the box with Aider, OpenCode, Cline, Cursor, Continue, Goose, Windsurf, and Claude Code.
Office Automation
Production-Ready Deliverables
Trained with senior professionals in finance, law, and social sciences. Word formatting, PowerPoint editing, Excel financial modeling, email drafting — ready from day one.
Measured on EU Infrastructure
MiniMax M2.5
10K input / 1K output, 1 concurrent, 10 requests
Up to 426 tok/s on shorter prompts. Last measured: April 2026.
10x Lower Cost Than Sonnet
Comparable SWE-bench performance at a fraction of the price.
| Model | Input (per 1M) | Output (per 1M) | Relative Cost |
|---|---|---|---|
| Claude Sonnet 4.6 | $3.00 | $15.00 | ~12x more |
| GPT-4o | $2.50 | $10.00 | ~8x more |
| MiniMax M2.5 (Infercom) | €0.30 | €1.20 | Baseline |
"MiniMax M2.5 is the first open-weight model to surpass Claude Sonnet on SWE-bench Verified."
— OpenHands
Don't Take Our Word For It
"M2.5 gave me the best result I've gotten so far. Better than Claude Code with Opus 4.6."
"I went from Claude Max ($100/month) to MiniMax ($20/month) with the same usage patterns and haven't hit limits once."
"There is a model that I can impartially say is basically up to the quality of Claude Sonnet."
"MiniMax 2.5 did a great job synthesizing research from multiple sources into an accurate table. I'd give it an A+ for developers."
Why EU Hosting Matters for MiniMax
MiniMax is a Chinese AI company. Running their model through Infercom means:
- Your data is processed in Germany, not China
- No US CLOUD Act exposure either
- Full GDPR compliance with EU-based DPA
- ISO 27001 certified infrastructure
- Zero data retention — we don't train on your data
Start Building in Minutes
from openai import OpenAI
client = OpenAI(
api_key="your-infercom-key",
base_url="https://api.infercom.ai/v1"
)
response = client.chat.completions.create(
model="MiniMax-M2.5",
messages=[{"role": "user", "content": "Your prompt here"}],
max_tokens=4096
)
print(response.choices[0].message.content)