LLM Reference
MiniMax

MiniMax

25 models across 7 families · Latest: MiniMax M3 (2026-06)

Developing AI for gaming and entertainment.

CodingRAGAgentsLong contextVisionClassificationJSON / Tool useChinaHighlight

MiniMax's portfolio covers 24 active models across 6 current families, spanning coding, rag, and agents. Open a model detail page to compare provider routes and sourced benchmarks.

Covers 7 workload areas across 24 active tracked models; last verified 2026-07-01.

Use it for

  • Teams evaluating coding, rag, and agents across this lab's releases
  • Comparing model families before committing to a flagship
  • Migration and pricing follow-ups across 24 tracked models

Do not use it for

  • Choosing a hosting provider without opening a model page for price ladders

Active models

24

Current models from this lab, excluding deprecated ones

Active families

6

Current model families from this lab

Open catalog

17 open

15 open source / 2 open weights

Lowest output price

$0.900 /1M

Cheapest tracked output across active models, per 1M tokens

Latest dated release

2026-06-01

MiniMax M3

Freshness

2026-07-01

Researched 1d ago

fresh

Information

Founded2021
Minhang, Shanghai, China

Release cadence

Showing 5 recent dated releases (full timeline below). Latest: MiniMax M3 (2026-06-01).

Where this lab wins

  • Coding: 4 tracked models with SWE-bench / HumanEval-style scores.
  • RAG: 7 tracked models with ruler / needle retrieval benchmarks.
  • Agentic: 4 tracked models with BFCL, tau-bench, and SWE-bench tool-use coverage.
  • Long-context: 14 tracked models with context-token or InfiniteBench-class signal.

Flagship quality / price signal

Flagship: MiniMax M3 (best sourced coding quality-per-dollar in this portfolio).

Coding task grade C · swe-bench-pro score 59 · cheapest tracked output $1.20 per 1M tokens

MiniMax is a Chinese AI research organization founded in 2021. Developing AI for gaming and entertainment. MiniMax ships 7 model families totaling 25 models, with the most recent release MiniMax M3 in 2026-06. Notable families include MiniMax M3, MiniMax M2, and MiniMax M1. Use it as a stable reference for lab background, release coverage, and follow-up model pages as they are added. View official API endpoints, benchmark performance, and coding/agent fit for every MiniMax model.

About

MiniMax is a Chinese AI company that specializes in developing AI technologies for the gaming and entertainment industries. Their work focuses on creating intelligent game agents, procedural content generation, and player behavior analysis to enhance gaming experiences and drive innovation. For example, MiniMax's AI can power non-player characters (NPCs) that exhibit more realistic and adaptive behaviors, making game worlds feel more immersive and dynamic. Their procedural content generation techniques can create endless variations of game levels, quests, and items, ensuring a fresh experience for players. By analyzing player behavior data, MiniMax's AI can help game developers optimize game balance, difficulty, and monetization strategies. Software engineers and SaaS executives in the gaming and entertainment industries can leverage MiniMax's technologies to build more engaging, personalized, and data-driven gaming experiences that captivate players and drive business success.

Featured models

ModelReleasedContextInput price ($/1M)Output price ($/1M)LicenseOpenness
MiniMax M32026-06-011m$0.30$1.20MiniMax Community LicenseOpen weights
MiniMax M2.72026-03-18205k$0.279$1.2MITOpen source
MiniMax M2.7 Highspeed2026-03-18205k$0.6$2.4MITOpen source

Model families

Recent releases

  1. MiniMax M3- 2026-06-01
  2. MiniMax M2.7- 2026-03-18
  3. MiniMax M2.7 Highspeed- 2026-03-18
  4. MiniMax M2-her- 2026-03-01
  5. Together MiniMax M2.5- 2026-02-15

Top comparisons

FAQ

Who founded MiniMax and when?

MiniMax was founded in 2021 and is associated with Minhang, Shanghai, China.

What models has MiniMax released?

MiniMax ships 25 models across 7 families: MiniMax M3, MiniMax M2, and MiniMax M1.

Is MiniMax's technology open source?

Some MiniMax models are open-weight (MiniMax M3, MiniMax M2.7, and MiniMax M2.7 Highspeed); others are proprietary (MiniMax Image-01, MiniMax Speech 2.8, and abab 6.5).

Where is MiniMax headquartered?

MiniMax is headquartered in Minhang, Shanghai, China.

What is MiniMax known for?

Developing AI for gaming and entertainment. Its most prominent tracked family is MiniMax M3.

How can I access MiniMax's models?

MiniMax's models are available via MiniMax, AWS Bedrock, Fireworks AI, Novita AI, and NVIDIA NIM.

Explore related pages

Last reviewed: 2026-07-01. Data sourced from public lab announcements and provider documentation.