LLMcompare

Search

Search for a command to run...

A · closed · Claude Sonnet / Haiku

Claude Sonnet 4.6

Anthropic

B · closed · Claude Sonnet / Haiku

Claude Sonnet 5

Anthropic

2:11

overall capability wins (1 ties across 14 scored). Claude Sonnet 5 leads overall.

  • Reasoning1:5
  • Coding0:4
  • Arena1:0
  • Tool use0:1
  • Composite indices0:2
  • Pricing0:3
  • Speed0:1
  • Specs0:1

Verdict

Claude Sonnet 5 leads coding (widest gap: +15.7 on WebDev Elo); Claude Sonnet 4.6 ranks higher on LMArena (+11 Elo); Claude Sonnet 5 allows ~2.0x more max output (128K vs 64K).

  • Claude Sonnet 5 leads coding (widest gap: +15.7 on WebDev Elo)
  • Claude Sonnet 4.6 ranks higher on LMArena (+11 Elo)
  • Claude Sonnet 5 allows ~2.0x more max output (128K vs 64K)
  • Claude Sonnet 5 is ~1.5x cheaper on a blended token basis than Claude Sonnet 4.6

Cost for 1M in + 250K out

Illustrative chat workload at primary-provider list prices.

Claude Sonnet 4.6
$6.75
Claude Sonnet 5
$4.50

Claude Sonnet 5 is about 1.5x cheaper on this workload.

Price per 1M tokens

Full comparison

Organization

Anthropic

Anthropic

Family

Claude Sonnet / Haiku

Claude Sonnet / Haiku

License

Proprietary

Proprietary

Open weights

No

No

Release

Mar 18, 2026

Jun 30, 2026

Knowledge cutoff

-

-

API / provider

Anthropic

Anthropic

Modalities

text, image → text

text, image → text

Specs

Context window

1M

1M

tie

Max output

64K

128K

B +64K

Parameters

—

—

—

Pricing

Input $/1M

$3

$2

B +$1

Output $/1M

$15

$10

B +$5

Blended $/1M (3∶1)

$6

$4

B +$2

Speed

tok/s

90

95

B +5.00

TTFT (s)

0.4

0.4

tie

Reasoning

MMLU-Pro

—

—

—

GPQA Diamond

87.5

91.1

B +3.6 pts

Humanity's Last Exam

33.6

41.3

B +7.7 pts

AIME 2025

—

—

—

MATH-500

—

—

—

Humanity's Last Exam (with tools)

—

—

—

AA-Omniscience Accuracy

40.9

40.1

A +0.8 pts

AA-LCR v1.1

80.0

82.0

B +2.0 pts

CritPt

3.1

16.9

B +13.8 pts

MMMU-Pro

73.3

77.3

B +4.0 pts

IFBench

56.6

—

—

Chartography

—

15.6

—

Chartography (With Tools)

—

—

—

Coding

SWE-bench Verified

—

—

—

SWE-bench Pro

—

—

—

SWE-bench Multilingual

—

—

—

LiveCodeBench

—

—

—

Terminal-Bench 2.1

71.2

80.5

B +9.3 pts

Aider Polyglot

—

—

—

Terminal-Bench 3

—

14.6

—

BigCodeBench

—

—

—

SciCode

50.1

54.3

B +4.2 pts

CursorBench

—

61.5

—

SWE-Rebench

54.5

56.8

B +2.3 pts

NL2Repo-Bench

—

—

—

DeepSWE

—

54.0

—

WebDev Arena

1,521

1,537

B +16 Elo

Terminal-Bench 4.0

—

12.4

—

CursorBench 4.0

—

—

—

FrontierCode v1.1 (Main)

—

—

—

Arena

LMArena Elo

1,473

1,461

A +11 Elo

Claude Sonnet 4.6

Context
1M / 64K out
Parameters
—
Price
$3 / $15
Speed
90 tok/s · 0.4s TTFT
Modalities
text, image → text
License
Proprietary

Prior Sonnet generation with strong coding and long-context support before Sonnet 5.

Claude Sonnet 5

Context
1M / 128K out
Parameters
—
Price
$2 / $10
Speed
95 tok/s · 0.4s TTFT
Modalities
text, image → text
License
Proprietary

Default Claude model for Free/Pro at $2/$10 (the launch introductory rate, made permanent on 2026-08-10 - the planned 2026-09-01 increase to $3/$15 was cancelled); closes much of the gap to prior Opus on agentic tasks.

Benchmark charts

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

Reasoning & knowledge

GPQA: B +3.6HLE: B +7.7CritPt: B +13.8AA-LCR: B +2.0Omniscience Acc.: A +0.8MMMU-Pro: B +4.0

Coding

TermBench: B +9.3SciCode: B +4.2SWE-Rebench: B +2.3WebDev Elo: B +16

Tool use & function calling

τ³-Banking: B +2.9

Arena

Arena Elo: A +11

Composite indices

AA Index: B +14.6AA Index v4.3: B +7.9

Pick a different pair · Back to catalog