← All models

MiniMax-M1-80k

Published Publisher MiniMax

Model positioning

MiniMax-M1-80k is a long-context reasoning model in the MiniMax-M1 series, designed for complex tasks requiring long inputs and deep thinking. The model uses a mixture-of-experts (MoE) architecture and a hybrid attention architecture incorporating Lightning Attention. It has 456 billion parameters in total, with approximately 45.9 billion activated per token. The official model card describes 80k as its thinking budget. (huggingface.co)

Capabilities and uses

  • Natively supports a context of approximately 1 million tokens and reasoning output of up to approximately 80,000 tokens. (minimax.io)
  • Suitable for long-document analysis, technical writing, complex problem solving, software engineering, and agent tasks requiring tool calls. (minimax.io)
  • At release, MiniMax reported that the 80k version scored 56.0% on the SWE-bench Verified set; benchmark results do not guarantee performance on real-world tasks. (minimax.cn)

Weights and limitations

MiniMax has released the model weights, and the model card specifies an Apache-2.0 license; use must comply with the license and deployment tool requirements. (huggingface.co)

Long contexts and deep reasoning require corresponding computational resources and time; model outputs may still contain errors, and important results should be verified. The context and reasoning output limits in the official announcements do not mean that all platforms or deployment configurations provide the same allowances.

Pricing

The following are the official website API prices for the M1 series published by MiniMax in its June 16, 2025 release announcements; they are not verified current prices. The announcements did not list a separate price or API identifier for the 80k version. The domestic and international prices apply to official services in different regions and should not be treated as the same plan with directly convertible prices. Date checked: November 18, 2026.

Domestic site: Chinese yuan (CNY)

Tiered by input length, with all prices per million tokens:

Input length Input price Output price
0–32k tokens 0.8 yuan 8 yuan
32k–128k tokens 1.2 yuan 16 yuan
128k–1M tokens 2.4 yuan 24 yuan

Source: MiniMax domestic release announcement. The announcement did not specify which tiers include the boundary values. (minimax.cn)

International site: US dollars (USD)

Tiered by input length, with all prices per million tokens:

Input length Input price Output price
0–200k tokens 0.4 USD 2.2 USD
200k–1M tokens 1.3 USD 2.2 USD

Source: MiniMax international release announcement. The announcement did not specify which tiers include the boundary values. (minimax.io)

Current applicable prices and availability have not been verified; the international site's current pricing overview does not list M1. Confirm with the relevant platform before use. (platform.minimax.io)

Release date and official sources

Release date: June 16, 2025. (minimax.io)