---
title: "GPT 5.4 mini Benchmarks - RankedAGI"
description: "Detailed benchmark and metadata record for GPT 5.4 mini."
source: "https://rankedagi.com/models/gpt-5-4-mini"
---

# GPT 5.4 mini

| Field | Value |
| --- | --- |
| Organization | OpenAI |
| License | Proprietary |
| Type | Thinking |
| Version | Latest |
| Released | 2026-03-17 |
| Context window | 400K |
| Knowledge cutoff | 2025-08-31 |
| Input cost per million tokens | $0.75 |
| Output cost per million tokens | $4.5 |
| Cached input cost per million tokens | $0.08 |
| Last updated | 2026-05-28T08:57:27.471Z |

## Benchmarks

| Benchmark | Category | Value | Description | Source |
| --- | --- | --- | --- | --- |
| RankedAGI Coding | coding | 68.0% | RankedAGI Coding Score |  |
| SWEBench Pro | coding | 54.4% | Diverse Agentic Coding |  |
| Terminal-Bench 2.0 | coding | 60.0% | Agentic Terminal Coding | https://www.tbench.ai/leaderboard/terminal-bench/2.0 |
| RankedAGI Agentic | Agents | 58.8% | RankedAGI Agentic Score |  |
| Code Arena | coding | 1397 | Code Arena Elo Score | https://arena.ai/leaderboard/code |
| Livebench Coding 26.1 | coding | 71.6% | Livebench Coding 26.01.08 | https://livebench.ai/#/?Coding=a&sort=Coding+Average&variants=true&highunseenbias=true |
| DeepSWE | coding | 24.0% | Measuring frontier coding agents on original, long-horizon engineering tasks | https://deepswe.datacurve.ai/ |
| LiveBench Agentic Coding 26.1 | coding | 47.5% |  | https://livebench.ai/#/?Agentic+Coding=a&sort=Agentic+Coding+Average&variants=true&highunseenbias=true |
| OSWorld-Verified | Agents | 72.1% |  |  |
| SvelteBench | coding | 83.3% | SvelteBench - Benchmark for Svelte | https://khromov.github.io/svelte-bench/benchmark-results-merged.html |
| RankedAGI Reasoning | reasoning | 66.0% | RankedAGI Reasoning Score |  |
| Humanity's Last Exam | reasoning | 28.2% | Multidisciplinary Reasoning (no tools) |  |
| Humanity's Last Exam w/ Tools | reasoning | 28.2% | Multidisciplinary Reasoning (with tools) |  |
| GPQA Diamond | reasoning | 88.0% | Generalized Prefix Question Answering Score (Reasoning) PhD Level Reasoning |  |
| Text Arena | general | 1456 | ChatArena (LMSYS) ELO Score | https://arena.ai/leaderboard/text |
| MMMU-Pro | imaging | 76.6% | Multimodal Understanding |  |
| MMMU Pro w/Tools | imaging | 78.0% |  |  |
| RankedAGI Math | math | 72.3% | RankedAGI Math Score |  |
| RankedAGI Overall | general | 60.8% | Overall RankedAGI score |  |
| 𝜏²-Bench Telecom | Agents | 93.4% | Agentic Tool Use |  |
| GDPval-AA Elo | Agents | 1427 | Office Tasks (Artificial Analysis) |  |
| MCP Atlas | Agents | 57.7% | Scaled Tool Use |  |
| Toolathlon | Agents | 42.9% | Benchmark to assess language agents' general tool use in realistic environments. | https://toolathlon.xyz/introduction |

## Model Links

| Title | URL |
| --- | --- |
| Release | https://openai.com/index/introducing-gpt-5-4-mini-and-nano/ |