Code with MiniMax M2.5 Highspeed
in AdaL CLI
Competitive coding capabilities at $0.30 per million input tokens — 10x cheaper than Sonnet, with open weights under MIT license. The model that makes agentic AI economical.
No credit card required
Why developers choose MiniMax M2.5
Maximum capability, minimal cost
10x Cheaper Than Sonnet
$0.30/1M input tokens vs $3.00 for Sonnet 4.6. Run massive agentic loops and codebase-wide refactors without budget anxiety.
~54 Tokens Per Second
Independently measured output speed of ~54 tok/s. Fast enough to feel like a real-time coding partner in your terminal.
Strong Multilingual Coding
Built with deep multilingual training. Excellent for internationalized codebases and non-English documentation.
Open Weights (MIT License)
230B parameter MoE model with only 10B active per token. Fully open weights under MIT license — self-host or use via API.
Competitive Reasoning
Scores 42 on the Artificial Analysis Intelligence Index, well above average among comparable open-weight models.
Built for Scale
When you need to rewrite hundreds of files or generate extensive test suites, MiniMax delivers quality at a price that lets you iterate freely.
MiniMax M2.5 Cost & Specs
10x cheaper than the competition
| Metric | MiniMax M2.5 | Sonnet 4.6 | Opus 4.6 | GPT-5.2 |
|---|---|---|---|---|
| Input (per 1M tokens) | $0.30 | $3.00 | $15.00 | ~$10.00 |
| Output (per 1M tokens) | $1.20 | $15.00 | $75.00 | ~$30.00 |
| Context Window | 205K | 200K | 200K | 400K |
| Output Speed | ~54 tok/s | — | — | — |
| Parameters (Active) | 10B (MoE) | — | — | — |
Source: Artificial Analysis — MiniMax M2.5 · Anthropic — Claude Opus 4.6
MiniMax M2.5 Coding Performance
Competitive with frontier models at 10x less cost
| Benchmark | MiniMax M2.5 | Sonnet 4.6 | Opus 4.6 | Gemini 3.1 Pro |
|---|---|---|---|---|
| SWE-bench Verified (Agentic coding) | 80.2% | 79.6% | 80.8% | 80.6% |
| Multi-SWE-bench (Multi-repo coding) | 51.3% | — | — | — |
| BrowseComp (Agentic search) | 76.3% | 74.7% | 84.0% | 85.9% |
Source: MiniMax Official Blog — M2.5 Release · Google DeepMind Model Card
How to use MiniMax M2.5 Highspeed
Get started in minutes
Install AdaL CLI
One command install, no Node.js needed. Works on macOS, Linux, and Windows.
Sign up (free)
Create your account — takes less than a minute.
Select MiniMax M2.5 Highspeed
Use /model to pick MiniMax M2.5 Highspeed from the model selector.
Ready to code without budget limits?
Experience competitive agentic coding at 10x less cost with MiniMax M2.5 Highspeed in AdaL CLI.
Get Started with AdaL