---
title: "Minimax M3 Benchmarks - RankedAGI"
description: "Detailed benchmark and metadata record for Minimax M3."
source: "https://rankedagi.com/models/minimax-m3"
---

# Minimax M3

| Field | Value |
| --- | --- |
| Organization | MiniMax |
| License | Open Source |
| Type | Thinking |
| Version | Latest |
| Released | 2026-06-01 |
| Input cost per million tokens | $0.6 |
| Output cost per million tokens | $2.4 |
| Cached input cost per million tokens | $0.12 |
| Last updated | 2026-06-17T19:32:55.052Z |

## Benchmarks

| Benchmark | Category | Value | Description | Source |
| --- | --- | --- | --- | --- |
| RankedAGI Coding | coding | 73.6% | RankedAGI Coding Score |  |
| SWEBench Pro | coding | 59.0% | Diverse Agentic Coding |  |
| SWEBench Verified | coding | 80.5% | Agentic Coding | https://www.swebench.com/index.html |
| Terminal-Bench 2.1 | coding | 66.0% | Agentic Terminal Coding | https://www.tbench.ai/leaderboard/terminal-bench/2.1 |
| RankedAGI Agentic | Agents | 67.7% | RankedAGI Agentic Score |  |
| BrowseComp | Agents | 83.5% | A benchmark for browsing agents | https://openai.com/index/browsecomp/ |
| DeepSWE | coding | 20.0% | Measuring frontier coding agents on original, long-horizon engineering tasks | https://deepswe.datacurve.ai/ |
| OSWorld-Verified | Agents | 70.0% |  |  |
| SvelteBench | coding | 96.7% | SvelteBench - Benchmark for Svelte | https://khromov.github.io/svelte-bench/benchmark-results-merged.html |
| RankedAGI Reasoning | reasoning | 71.8% | RankedAGI Reasoning Score |  |
| MMMU-Pro | imaging | 78.1% | Multimodal Understanding |  |
| RankedAGI Math | math | 69.5% | RankedAGI Math Score |  |
| RankedAGI Overall | general | 64.8% | Overall RankedAGI score |  |
| MCP Atlas | Agents | 74.2% | Scaled Tool Use |  |
| Code DesignArena | Design | 1308 |  | https://www.designarena.ai/leaderboard/code |

## Model Links

| Title | URL |
| --- | --- |
| Blog | https://www.minimax.io/blog/minimax-m3 |