Skip to content
Leaderboard / OpenAI

GPT-5.4 mini

OpenAI High CLOSED Proprietary
Verdict

GPT-5.4 mini is a solid mid-pack model: #149 of 406 by SIscope score. Strongest at following instructions, weaker at coding. Mid-range on price.

SIscope score
57
#149 of 406
Human-vote rank
#139
of 419 · score 1412
Price per 1M tokens, blended
$1.69
$0.75 in · $4.5 out
Context window
400K
400,000 tokens

Shape

Each spoke is a category, 0 at the centre, 100 at the edge.

CodingAgentsno data yetReasoningMathChatInstructions

Category scores

Percentiles: 90 means better than about 90% of tested models.

Coding52
Agents & toolsNo data yet–
Reasoning & knowledge53
Math69
Chat quality67
Following instructions72

What goes into each category · Compare with DeepSeek V4.1 Flash

Human-vote score over time

Text arena, recent updates.

140514131420Oct 20: 1411.6 (rank #138)Oct 22: 1411.5 (rank #138)Oct 28: 1411.6 (rank #139)Oct 20Oct 28

Coding votes over time

WebDev arena, recent updates.

139013981405Oct 16: 1397.7 (rank #95)Oct 18: 1397.7 (rank #96)Oct 19: 1397.4 (rank #99)Oct 22: 1397.4 (rank #100)Oct 23: 1397.5 (rank #101)Oct 27: 1397.1 (rank #102)Oct 28: 1397.3 (rank #103)Oct 16Oct 28

Independent test runs

Run by Epoch AI on the same harness for every model. Bars show the share of questions answered correctly.

GPQA Diamond84%
PhD-level science multiple choice questions
Better than 66% of 333 tested models. Run Apr 15, 2026.
SimpleQA Verified29%
Short factual questions; rewards not making things up
Better than 23% of 100 tested models. Run Aug 27, 2026.
OTIS Mock AIME 2024-2587%
Competition maths problems in the style of AIME
Better than 70% of 313 tested models. Run Apr 15, 2026.

Human-vote arenas

People compare two anonymous answers and pick the better one. Scores work like chess ratings.

ArenaRankScoreRange (95%)VotesUpdated
Text arena
Everyday chat, blind human votes
#139 of 41914121408–141563,432Oct 28, 2026
WebDev arena
Building web apps, blind human votes
#103 of 15213971391–140411,893Oct 28, 2026
Vision arena
Questions about images, blind human votes
#69 of 16712451239–125125,734Oct 19, 2026
Agent arena
Real multi-step agent sessions
Not on this board yet

Source: Arena (formerly LMArena) leaderboard dataset, CC BY 4.0. Fetched Nov 4, 2026.

Price

Input
$0.75
per 1M tokens
Output
$4.5
per 1M tokens
Blended
$1.69
3 in : 1 out, per 1M
Context window
400K
max tokens per request
Speed
Coming soon
our own tests

List price from the OpenRouter models API (openai/gpt-5.4-mini), Nov 4, 2026. Providers can charge differently.