Compare

Four models, one page

Pick the models you are actually choosing between. Every row is scaled within itself, so a bar means something next to its neighbours and nothing across rows.

Slot 1

Slot 2

Slot 3

Slot 4

Specification	GPT-5.4 MiniOpenAI	Claude Opus 5Anthropic
Released	Mar 17, 2026	Jul 24, 2026
Context window	400K	1M
Max output	128K	128K
Input / 1M	$0.75	$5
Output / 1M	$4.5	$25
Cost per task	$0.40	$2.25
Arena Elo	—	1,393
Serving providers	2	5
Parameters	Undisclosed	Undisclosed
Licence	Proprietary	Proprietary
Capabilities	Reasoning Tool use Vision	Reasoning Tool use Vision

Metrics side by side

Each row is scaled to the largest value in that row — bars compare within a row, never across rows

GPT-5.4 Mini
Claude Opus 5

Intelligence Index
GPT-5.4 Mini
40.9
Claude Opus 5
63.1
Coding Index
GPT-5.4 Mini
56.1
Claude Opus 5
78.0
Agentic Index
GPT-5.4 Mini
31.5
Claude Opus 5
59.2
Output speed
GPT-5.4 Mini
104 t/s
Claude Opus 5
87 t/s
Context window
GPT-5.4 Mini
400K
Claude Opus 5
1M
Latency · lower is better
GPT-5.4 Mini
828ms
Claude Opus 5
1.12s
Blended price / 1M · lower is better
GPT-5.4 Mini
$1.69
Claude Opus 5
$10
Cost per task · lower is better
GPT-5.4 Mini
$0.40
Claude Opus 5
$2.25

Rows marked “lower is better” still draw a longer bar for a larger number — read the value, not just the length. Arena Elo is in the specification table above instead: it has no meaningful zero, so a bar would flatten the gaps.

View as table

Metric	GPT-5.4 Mini	Claude Opus 5
Intelligence Index	40.9	63.1
Coding Index	56.1	78.0
Agentic Index	31.5	59.2
Output speed	104 t/s	87 t/s
Context window	400K	1M
Latency · lower is better	828ms	1.12s
Blended price / 1M · lower is better	$1.69	$10
Cost per task · lower is better	$0.40	$2.25

Evaluation scores

Percentage correct on a common 0–100% scale

GPT-5.4 Mini
Claude Opus 5

GPQA Diamond
GPT-5.4 Mini
87.5%
Claude Opus 5
93.2%
Humanity's Last Exam
GPT-5.4 Mini
28.1%
Claude Opus 5
54.9%
SciCode
GPT-5.4 Mini
49.9%
Claude Opus 5
55.7%
τ²-bench
GPT-5.4 Mini
83.3%
Claude Opus 5
42.1%
Terminal-Bench Hard
GPT-5.4 Mini
52.3%
Claude Opus 5not measured
AA-LCR long context
GPT-5.4 Mini
73.0%
Claude Opus 5
75.7%

A missing bar means that evaluation was not run for that model — it is not a zero.

View as table

Evaluation	GPT-5.4 Mini	Claude Opus 5
GPQA Diamond	87.5%	93.2%
Humanity's Last Exam	28.1%	54.9%
SciCode	49.9%	55.7%
τ²-bench	83.3%	42.1%
Terminal-Bench Hard	52.3%	—
LiveCodeBench	—	—
AA-LCR long context	73.0%	75.7%
AIME 2025	—	—

Estimated from list pricing: 50K input tokens plus 80K output tokens for reasoning models (25K for non-reasoning).