Compare Models

GPT 5.4 Mini

GPT 5.4 Mini is a model from OpenAI, released March 17, 2026.

Release Date: Mar 17, 2026

Developer OpenAI  🇺🇸
Context Window 400k
Max Output Tokens 128k
Token Costs (in/out) $0.75/4.50
Weights Private
Input Modalities

Accuracy

47.72 %

Avg. Cost (In/Out)

$ 0.75 / $ 4.50

Latency

20 min 8 s

Vals Index
BenchmarksAccuracyRankings
Code Migration

12.94%

±3.64
62/79
EMB

45.43%

±3.65
58/76
Finance Agent (v2)

45.36%

±0.45
56/81
Legal Research Bench

12.50%

±2.30
77/89
MortgageTax

63.51%

±0.91
52/98
Tax Agent Bench

8.81%

±2.04
68/80
TaxEval v2

71.22%

±0.90
81/145
Vibe Code Bench v1.1

47.97%

±5.61
69/123
GPQA Diamond

83.08%

±2.46
59/138
LiveCodeBench

81.47%

±1.09
62/143
MMLU Pro

84.55%

±0.36
63/138
MMMU Pro

79.25%

±0.97
49/93
ProgramBench

0.00%

±0.00
43/64
SWE-bench

73.00%

±1.99
53/88
Terminal-Bench 4.0

2.52%

±1.01
53/59
Proprietary BenchmarksAcademic BenchmarksIndustry Partners
Vals
Default Provider : OpenAI
Temperature: Default
Top P: Default
Top K: Default
Max Output Tokens: 128,000
Reasoning Effort: xhigh

Compared with

GPT 5.4 Mini vs Gemini 3.1 Pro Preview (02/26) 12 shared benchmarks GPT 5.4 Mini vs MiMo V2.5 Pro 9 shared benchmarks GPT 5.4 Mini vs Inkling 12 shared benchmarks GPT 5.4 Mini vs Inkling Small 11 shared benchmarks GPT 5.4 Mini vs MiniMax-M3 11 shared benchmarks GPT 5.4 Mini vs Mercury 2.5 4 shared benchmarks GPT 5.4 Mini vs DeepSeek V4 10 shared benchmarks GPT 5.4 Mini vs GLM 5.3 Flash 11 shared benchmarks

Updates

Mar 17, 2026

We evaluated GPT 5.4 Mini across our benchmarks.

All evaluations were run using the official OpenAI API using “xhigh” reasoning, except on Terminal-Bench 2.0, where we used “high”. The numbers on the Vals Index are finalized, but on certain benchmarks, we are still confirming results based on the latest parameters.