# MiniMax Brand Voice

> A high-performance, engineering-centric voice that balances technical precision with collaborative utility.

## Positioning
MiniMax is a global leader in general artificial intelligence, providing frontier models (M3) and AI-native products for coding, multimodal generation, and agentic workflows. It targets developers and enterprises who require "engineering-grade" performance, massive 1M context windows, and scalable intelligence.

## Voice principles
*   **Technical & Precise:** Uses specific architectural terms (MSA, Sparse Attention, Interleaved Data) to establish authority without unnecessary fluff.
*   **Performance-Driven:** Focuses on scale and capability, using words like "Frontier," "Full-blooded (满血)," and "Top-tier (顶尖)."
*   **Collaborative:** Positions the AI as a partner rather than just a tool, emphasizing "co-creation," "teamwork," and "tacit understanding."
*   **Direct & Comparative:** Uses blunt, data-driven comparisons to define value, especially regarding pricing and capacity.

## Tone by context
| Context | Tone |
|---|---|
| Product Headings | Bold, punchy, and feature-forward. Focuses on "Frontier" capabilities. |
| Technical Research | Academic and rigorous. Uses industry-standard terminology for AI architecture. |
| Pricing (Token Plan) | Pragmatic and value-oriented. Uses direct price-to-value comparisons. |
| Mission Statements | Visionary but grounded. Focuses on universal access and "co-creation." |

## Lexicon
- **Use:** 前沿 (Frontier/Leading-edge), 真实工程级 (Real engineering-grade), 满血 (Full-blooded/Max-spec), 默契 (Tacit understanding/Rapport), 沉淀 (Accumulate/Settle), 召集 (Summon/Assemble), 原生 (Native).
- **Avoid:** Not evident from captured copy (the brand avoids flowery marketing metaphors in favor of technical specs).

## Messaging do's and don'ts
*   **Do:** Emphasize the "1M context" and "MSA architecture" as unique technical advantages.
*   **Do:** Use "Agent" terminology to describe task-solving rather than just "Chatbot."
*   **Do:** Highlight the transition from "generating code" to "understanding collaboration."
*   **Don't:** Use vague superlatives; back up claims with version numbers (Speech 2.8) or specific metrics (360,000 documents).
*   **Don't:** Hide the complexity of the tech; lean into the "Step 0" and "Sparse Attention" details.

## Evidence
- "真实工程级，不止生成代码，更懂协作" (Engineering-grade, collaborative focus).
- "全新稀疏注意力，让 context 真正可 scale" (Technical precision, MSA architecture).
- "Token Plan 满血升级 · 用量翻倍" (Direct, performance-oriented marketing).
- "与所有人共创智能" (Mission-driven, collaborative).
- "Max ¥119 ≈ 36 万份长文档" (Concrete, comparative value).