LLMcompare

Search

Search for a command to run...

A · closed · Claude Opus

Claude Opus 4.7

Anthropic

B · closed · GPT-5.x

GPT-5.4

OpenAI

5:10

overall capability wins (1 ties across 16 scored). GPT-5.4 leads overall.

  • Reasoning1:6
  • Coding3:2
  • Arena1:0
  • Tool use0:1
  • Pricing0:3
  • Speed0:2
  • Specs0:1

Verdict

Claude Opus 4.7 leads coding (widest gap: +170.7 on WebDev Elo); GPT-5.4 edges reasoning & knowledge; Claude Opus 4.7 ranks higher on LMArena (+29 Elo).

  • Claude Opus 4.7 leads coding (widest gap: +170.7 on WebDev Elo)
  • GPT-5.4 edges reasoning & knowledge
  • Claude Opus 4.7 ranks higher on LMArena (+29 Elo)
  • GPT-5.4 offers larger context (1.1M vs 1M)
  • GPT-5.4 is ~1.8x cheaper on a blended token basis than Claude Opus 4.7

Cost for 1M in + 250K out

Illustrative chat workload at primary-provider list prices.

Claude Opus 4.7
$11.25
GPT-5.4
$6.25

GPT-5.4 is about 1.8x cheaper on this workload.

Price per 1M tokens

Full comparison

Organization

Anthropic

OpenAI

Family

Claude Opus

GPT-5.x

License

Proprietary

Proprietary

Open weights

No

No

Release

Apr 16, 2026

Mar 5, 2026

Knowledge cutoff

-

-

API / provider

Anthropic

OpenAI

Modalities

text, image → text

text, image → text

Specs

Context window

1M

1.1M

B +50K

Max output

128K

128K

tie

Parameters

—

—

—

Pricing

Input $/1M

$5

$2.50

B +$2.50

Output $/1M

$25

$15

B +$10

Blended $/1M (3∶1)

$10

$5.63

B +$4.38

Speed

tok/s

68

85

B +17

TTFT (s)

0.55

0.4

B +0.15

Reasoning

MMLU-Pro

—

—

—

GPQA Diamond

91.4

92.0

B +0.6 pts

Humanity's Last Exam

42.3

43.7

B +1.4 pts

AIME 2025

—

—

—

MATH-500

—

—

—

Humanity's Last Exam (with tools)

—

—

—

AA-Omniscience Accuracy

48.9

50.8

B +1.9 pts

AA-LCR v1.1

78.7

82.0

B +3.3 pts

CritPt

12.0

23.4

B +11.4 pts

MMMU-Pro

78.8

78.4

A +0.4 pts

IFBench

58.6

73.9

B +15.3 pts

Chartography

—

—

—

Chartography (With Tools)

—

—

—

Coding

SWE-bench Verified

—

—

—

SWE-bench Pro

64.3

57.7

A +6.6 pts

SWE-bench Multilingual

—

—

—

LiveCodeBench

—

—

—

Terminal-Bench 2.1

83.1

78.3

A +4.8 pts

Aider Polyglot

—

—

—

Terminal-Bench 3

—

—

—

BigCodeBench

—

—

—

SciCode

54.5

56.6

B +2.1 pts

CursorBench

—

—

—

SWE-Rebench

53.1

57.6

B +4.5 pts

NL2Repo-Bench

—

—

—

DeepSWE

—

—

—

WebDev Arena

1,557

1,387

A +171 Elo

Terminal-Bench 4.0

—

—

—

CursorBench 4.0

—

—

—

FrontierCode v1.1 (Main)

—

—

—

Arena

LMArena Elo

1,494

1,466

A +29 Elo

Claude Opus 4.7

Context
1M / 128K out
Parameters
—
Price
$5 / $25
Speed
68 tok/s · 0.55s TTFT
Modalities
text, image → text
License
Proprietary

April 2026 Opus release with a large SWE-bench jump; still widely used alongside Opus 4.8 / 5.

GPT-5.4

Context
1.1M / 128K out
Parameters
—
Price
$2.50 / $15
Speed
85 tok/s · 0.4s TTFT
Modalities
text, image → text
License
Proprietary

Solid mid-2026 OpenAI workhorse between GPT-5 and the 5.5/5.6 flagships.

Benchmark charts

Winner bars are emphasized. Per-benchmark deltas sit above each chart.

Reasoning & knowledge

GPQA: B +0.6HLE: B +1.4CritPt: B +11.4AA-LCR: B +3.3Omniscience Acc.: B +1.9MMMU-Pro: A +0.4IFBench: B +15.3

Coding

SWE-Pro: A +6.6TermBench: A +4.8SciCode: B +2.1SWE-Rebench: B +4.5WebDev Elo: A +171

Tool use & function calling

τ³-Banking: B +5.0

Arena

Arena Elo: A +29

Pick a different pair · Back to catalog