Model
Model explorer

GPT-5.4 mini

CLOSED
OpenAI · GPT-5.4 family · released Mar 17, 2026

Mid-tier sibling of GPT-5.4, launched Mar 17 2026 (12 days after the flagship/Pro) via OpenAI's own 'Introducing GPT-5.4 mini and nano' announcement as a distinct, separately-priced, separately-benchmarked model -- not a rename or repricing of GPT-5 mini. Runs >2x faster than GPT-5 mini and beats it on every published coding/tool/knowledge/vision benchmark; available in ChatGPT (Free/Go default, fallback for others), Codex, and the API.

ReasoningCodingVisionFunction callingTool useAgentic
2098.8
Elo · rank #52
Parameters
Undisclosed
Active params
—
Context
400K tokens
Architecture
Undisclosed architecture (presumed dense transformer); cost-efficient mid-tier sibling of GPT-5.4, optimized for coding, computer use, and subagent workloads, running more than 2x faster than the flagship
License
Proprietary (OpenAI API Terms)
Languages
—
API price (in/out)
$0.75 / $4.5
Modalities
text · vision
Benchmark results
Bar shows position within the tracked field; marker = field best
AIMEMath87.2%#57
best: GPT-5.2 · 100.0%
ARC-AGI-2Reasoning18.9%#18
best: GPT-5.6 · 92.5%
FrontierMath (Tiers 1-3)Math28.3%#9
best: GPT-5.6 · 89.0%
FrontierMath (Tier 4)Math2.1%#9
best: GPT-5.6 · 83.0%
GPQA DiamondReasoning88.0%#38
best: GPT-5.6 · 94.6%
Humanity's Last ExamReasoning28.2%#49
best: Claude Sonnet 5 · 57.4%
LiveCodeBenchCoding81.5%#42
best: DeepSeek-V4-Pro (Think Max) · 93.5%
MCP AtlasAgents57.7%#6
best: Kimi K3 · 84.2%
MMMU-ProVision76.6%#22
best: Claude Opus 4.7 · 85.5%
OSWorld-VerifiedAgents72.1%#14
best: Claude Fable 5 · 85.0%
SWE-bench ProCoding54.4%#27
best: Claude Fable 5 · 80.0%
SWE-bench VerifiedCoding73.0%#50
best: Claude Fable 5 · 95.0%
τ²-Bench TelecomAgents93.4%#16
best: Claude Opus 4.6 · 99.3%
Terminal-Bench 2.0Coding60.0%#30
best: GPT-5.6 · 88.8%
Run it locally
Closed weights — available via API only. No local deployment.
Input / M tok
$0.75
Output / M tok
$4.5
GPT-5.4 family
Elo progression across releases
GPT-5.4Mar 20262473.4GPT-5.4 ProMar 20262786.0GPT-5.4 miniMar 20262098.8GPT-5.4 nanoMar 20261965.6
Resources
DOCShttps://developers.openai.com/api/docs/models/gpt-5.4-miniAPIAvailable via OpenAI API
API price $0.75/$4.5 · each benchmark row carries its own source badge (see methodology)