Skip to content
Leaderboard / OpenAI

GPT-5.4

OpenAI High, Codex Harness CLOSED Proprietary
Other settings:High · 88Default · 67Medium, Codex Harness
Verdict

GPT-5.4 doesn't have enough test coverage for an overall SIscope score yet. Strongest at math, weaker at coding. On the pricey side: costs more than 84% of priced models.

SIscope score
No score yet
60% coverage, needs 70%
Human-vote rank
Not on board
Text arena
Price per 1M tokens, blended
$5.63
$2.5 in · $15 out
Context window
1.1M
1,050,000 tokens

Shape

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

CodingAgentsno data yetReasoningMathChatno data yetInstructionsno data yet

Category scores

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

Coding65
Agents & toolsNo data yet–
Reasoning & knowledge84
Math88
Chat qualityNo data yet–
Following instructionsNo data yet–

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

Coding votes over time

WebDev arena, recent updates.

145514631470Oct 16: 1463.8 (rank #69)Oct 18: 1464 (rank #70)Oct 19: 1463.9 (rank #72)Oct 22: 1463.4 (rank #73)Oct 23: 1463.7 (rank #74)Oct 27: 1463.3 (rank #75)Oct 28: 1463.5 (rank #76)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 Diamond90%
PhD-level science multiple choice questions
Better than 84% of 333 tested models. Run Jul 15, 2026.
OTIS Mock AIME 2024-2598%
Competition maths problems in the style of AIME
Better than 88% of 313 tested models. Run Jul 15, 2026.
SWE-bench Verified77%
Fixing real GitHub issues in Python repositories
Better than 79% of 35 tested models. Run Mar 6, 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
Not on this board yet
WebDev arena
Building web apps, blind human votes
#76 of 15214641445–14821,345Oct 28, 2026
Vision arena
Questions about images, blind human votes
Not on this board yet
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
$2.5
per 1M tokens
Output
$15
per 1M tokens
Blended
$5.63
3 in : 1 out, per 1M
Context window
1.1M
max tokens per request
Speed
Coming soon
our own tests

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