Skip to main content
BenchLM
Data

Claude Sonnet 4.5 vs GPT-5.4

Updated October 17, 2026. Rank says GPT-5.4 is ahead. Price, access, and your workload can each overturn that. Public scores include evidence status and uncertainty.

Share or export
Share on XLinkedInAPI/MCP

Decision reading

GPT-5.4 has the higher public point estimate, 66.7 versus 47.96. Their conditional score ranges do not overlap. These ranges do not establish rank confidence. 9 results are shared. Category rows resting on Estimated evidence or different benchmark sets are marked directional and do not name a winner.

Model A
Anthropic logo
Claude Sonnet 4.5

Anthropic

47.96/100

Estimated · Public rank #106

Conditional range 38.2–57.7

Model B
OpenAI logo
GPT-5.4

OpenAI

66.7/100

Supported · Public rank #30

90% interval 62.2–71.2

Shared results
9
Claude Sonnet 4.5 only
3
GPT-5.4 only
33
Like-for-like categories
0 / 8
Estimated: Claude Sonnet 4.5 · Supported: GPT-5.4. Conditional ranges do not establish rank confidence.How the comparison works

Which one for your work

Recommendations appear only when a shared evidence basis or an explicit operating constraint supports the call. Secondary and unsupported use cases stay disclosed below the initial list.

  • Long documents

    Prompts that approach the documented context limit

    GPT-5.4

    GPT-5.4 has the larger documented context window.

    Confidence: documented

  • Chat turn cost

    1K fresh input + 500 output tokens

    GPT-5.4

    GPT-5.4 has the lower estimated token cost for this stated workload. Costs use the listed standard API rates.

    Confidence: listed-rates
Show secondary and unsupported calls

  • Repository review cost

    50K fresh input + 3K output tokens

    GPT-5.4

    GPT-5.4 has the lower estimated token cost for this stated workload. Costs use the listed standard API rates.

    Confidence: listed-rates

  • Coding work

    Code generation, repair, and software-engineering tasks

    Not enough matched evidence

    Claude Sonnet 4.5 is not ranked on the public lane for coding, so no winner is named for coding.

    Confidence: limited

  • Agentic work

    Tool use, computer use, and multi-step task completion

    Not enough matched evidence

    Claude Sonnet 4.5 is scored on Estimated evidence for agentic, so the reading is directional rather than like-for-like.

    Confidence: limited

  • Cache-heavy agent loop cost

    200K cached + 20K fresh input + 10K output tokens

    Not enough matched evidence

    The page does not recommend a cost winner because at least one model cannot fit the stated workload in one request. Claude Sonnet 4.5 does not fit this workload in one request. Claude Sonnet 4.5 has no published cached-input rate, so cached tokens use its listed input rate.

    Confidence: rate-fallback

Which one for a specific job

Choose a job from the LLM Selector's task catalog to see the category row it rests on, under the same basis rules as the table below. A directional row stays directional; choosing a job never creates a winner.

The same task catalog as the LLM Selector. Each job names the evidence surface it rests on; nothing here adds a new score.

—Claude Sonnet 4.547.7GPT-5.4

Not comparable · BenchAlign v5.8

The coding row is not comparable on the public lane: at least one model is not measured or not ranked there.

Coding scores combine specific tasks and setups. Match the editor, harness, and effort to your workflow.

Open the coding leaderboardShortlist models for this job
Same basis rules as the category table below

What is actually comparable

Shared results can support a head-to-head reading. Results present for only one model describe coverage, not superiority.

1 category rests on Estimated evidence or different evidence sets. Those rows remain visible for coverage context but do not name a winner.

Shape of the matched evidence

Only shared public evidence is shown. Sparse evidence stays a ruled list rather than being closed into a radar shape.

Too few matched category axes support a radar. The ruled list below shows only shared benchmark results; positions use each benchmark’s normalized display scale when available.

Bars run 0–100 on each benchmark’s normalized display scale

Category results, on a stated basis

Each row shows the public-lane category score for both models: the BenchAlign v5.8 lane for agentic, coding, and knowledge, and the weighted public rows elsewhere, with the same rank each model page reports. A row is like-for-like only when both scores rest on Supported evidence or the same weighted set. Directional and not-comparable rows remain visible, but they never receive a winner.

Agentic

Directional only
Claude Sonnet 4.5
35.1
Estimated · #81/128
GPT-5.4
52.1
Supported · #44/128
Basis
BenchAlign v5.8 lane · 5 vs 15 public rows
Reading
Directional only

Coding

Not comparable
Claude Sonnet 4.5
Not ranked
GPT-5.4
47.7
Supported · #48/150
Basis
BenchAlign v5.8 lane · 2 vs 6 public rows
Reading
Not comparable

Reasoning

Not comparable
Claude Sonnet 4.5
22.9
Unranked · 1 rankable row
GPT-5.4
64.2
#23/28
Basis
Provisional lane · 1 vs 2 weighted rows
Reading
Not comparable

Multimodal

Not comparable
Claude Sonnet 4.5
Not ranked
GPT-5.4
73.1
#30/54
Basis
Provisional lane · 0 vs 3 weighted rows
Reading
Not comparable

Knowledge

Not comparable
Claude Sonnet 4.5
Not ranked
GPT-5.4
69.1
Supported · #20/184
Basis
BenchAlign v5.8 lane · 1 vs 7 public rows
Reading
Not comparable

Multilingual

Not comparable
Claude Sonnet 4.5
Not ranked
GPT-5.4
Not ranked
Basis
Provisional lane · 0 vs 0 weighted rows
Reading
Not comparable

Instruction following

Not comparable
Claude Sonnet 4.5
Not ranked
GPT-5.4
90.7
#23/148
Basis
Provisional lane · 0 vs 0 weighted rows
Reading
Not comparable

Math

Not comparable
Claude Sonnet 4.5
36.4
Unranked · 2 rankable rows
GPT-5.4
82.0
Unranked · 2 rankable rows
Basis
Provisional lane · 2 vs 2 weighted rows
Reading
Not comparable

Ranks count the models scored in each category’s lane, so the agentic, coding, and knowledge denominators (BenchAlign v5.8) differ from the provisional-lane categories. Unranked scores sit on the provisional lane’s scale but never name a winner. Like-for-like rows within 0.5 points read as a practical tie.

Supported evidence per lane · bars run 0–100Methodology

What each workload costs

Three fixed token mixes turn per-token rates into comparable decisions. Each scenario states context fit and whether cached input had to fall back to the published list-input rate.

Chat turn

1K fresh input + 500 output tokens

Claude Sonnet 4.5
$0.0105
Fits in one request
GPT-5.4
$0.01
Fits in one request

GPT-5.4 has the lower modeled cost

Costs use the listed standard API rates.

Repository review

50K fresh input + 3K output tokens

Claude Sonnet 4.5
$0.195
Fits in one request
GPT-5.4
$0.17
Fits in one request

GPT-5.4 has the lower modeled cost

Costs use the listed standard API rates.

Cache-heavy agent loop

200K cached + 20K fresh input + 10K output tokens

Claude Sonnet 4.5
$0.81
Does not fit in one request
Cached input priced at the published list-input rate
GPT-5.4
$0.25
Fits in one request

Claude Sonnet 4.5 does not fit this workload in one request. Claude Sonnet 4.5 has no published cached-input rate, so cached tokens use its listed input rate.

Cached input falls back to the list input rate only where a cached rate is unpublished

Specification differences

Sourced differences are shown directly. Missing facts stay explicit instead of being inferred from a model name or family.

Context window

Maximum documented context; output-token limits may be lower.

Claude Sonnet 4.5

200K

GPT-5.4

1.05M

OpenAI pricing

Cached-input rate

A missing cached-input rate falls back to the listed input rate only in the stated workload estimate.

Claude Sonnet 4.5

Not published

GPT-5.4

$0.25 per 1M cached input tokens

OpenAI pricing

Documented inputs

Claude Sonnet 4.5

Not sourced

GPT-5.4

Not sourced

Documented outputs

Claude Sonnet 4.5

Not sourced

GPT-5.4

Not sourced

Provider availability

Claude Sonnet 4.5

Not sourced

GPT-5.4

Not sourced

Reasoning profile

Claude Sonnet 4.5

Non-Reasoning

GPT-5.4

Reasoning

Weight access

Claude Sonnet 4.5

Proprietary

GPT-5.4

Proprietary

License

Claude Sonnet 4.5

Proprietary

GPT-5.4

Proprietary

Release date

Claude Sonnet 4.5

2025-09-29

GPT-5.4

2026-03-05

If you already use one of these models

Deployment change
The models list different providers, so authentication, endpoint behavior, limits, and feature support may change.
Quality signal
GPT-5.4 has the higher public point estimate, 66.7 versus 47.96. Their conditional score ranges do not overlap. These ranges do not establish rank confidence.
Workload cost
Repository review: $0.195 vs $0.17. Cache-heavy agent loop: $0.81 vs $0.25.
Context tradeoff
GPT-5.4 has the larger documented window (1.05M).
Run the same representative tasks against both endpoints before changing production traffic.

Questions

Which is better, Claude Sonnet 4.5 or GPT-5.4?

GPT-5.4 has the higher public point estimate, 66.7 versus 47.96. Their conditional score ranges do not overlap. These ranges do not establish rank confidence. The page therefore keeps the decision tied to the specific documented workload.

Which is better for coding, Claude Sonnet 4.5 or GPT-5.4?

Claude Sonnet 4.5 is not ranked on the public lane for coding, so no winner is named for coding.

Which is better for agentic tasks, Claude Sonnet 4.5 or GPT-5.4?

GPT-5.4 scores higher for agentic tasks on the public lane, 52.1 to 35.1. Claude Sonnet 4.5 is scored on Estimated evidence for agentic tasks, so the reading is directional rather than like-for-like. BenchLM does not name a winner for agentic tasks from a directional reading; read the shared benchmark rows directly and test the models on the same task set.

Which costs less, Claude Sonnet 4.5 or GPT-5.4?

For the stated presets, chat costs $0.0105 on Claude Sonnet 4.5 and $0.01 on GPT-5.4; repository review costs $0.195 and $0.17; the cache-heavy agent loop costs $0.81 and $0.25. Claude Sonnet 4.5 does not fit this workload in one request. Claude Sonnet 4.5 has no published cached-input rate, so cached tokens use its listed input rate.

Which has the larger context window, Claude Sonnet 4.5 or GPT-5.4?

GPT-5.4 has the larger documented context window: 1.05M, compared with 200K.

Benchmark evidence

The full public result ledger is available for audit without forcing a wide desktop table onto a phone.

Browse raw public benchmark evidence45 rows

Agentic

  • Terminal-Bench 2.0

    Claude Sonnet 4.550%
    Source
    GPT-5.475.1%
    Source

    GPT-5.4 leads this result

  • OSWorld-Verified

    Claude Sonnet 4.561.4%
    Source
    GPT-5.475%
    Source

    GPT-5.4 leads this result

  • VITA-Bench

    Claude Sonnet 4.517.0%
    Source
    GPT-5.4—

    Not directly comparable

  • Gert Labs

    Shared source
    Claude Sonnet 4.548.51%
    GPT-5.464.89%

    GPT-5.4 leads this result

  • JobBench

    Shared source
    Claude Sonnet 4.527.7%
    GPT-5.438.9%

    GPT-5.4 leads this result

  • CyberGym

    Claude Sonnet 4.5—
    GPT-5.479.0%
    Source

    Not directly comparable

  • BrowseComp

    Claude Sonnet 4.5—
    GPT-5.482.7%
    Source

    Not directly comparable

  • MCP Atlas

    Claude Sonnet 4.5—
    GPT-5.470.6%
    Source

    Not directly comparable

  • Toolathlon

    Claude Sonnet 4.5—
    GPT-5.454.6%
    Source

    Not directly comparable

  • τ²-bench results

    Claude Sonnet 4.5—
    GPT-5.498.9%
    Source

    Not directly comparable

  • Claw-Eval

    Claude Sonnet 4.5—
    GPT-5.460.3%
    Source

    Not directly comparable

  • DeepSearchQA

    Claude Sonnet 4.5—
    GPT-5.473.6%
    Source

    Not directly comparable

  • ResearchClawBench

    Claude Sonnet 4.5—
    GPT-5.415.3%
    Source

    Not directly comparable

  • ExploitGym

    Claude Sonnet 4.5—
    GPT-5.46.0%
    Source

    Not directly comparable

  • ApprenticeBench

    Claude Sonnet 4.5—
    GPT-5.411%
    Source

    Not directly comparable

  • MazeBench

    Claude Sonnet 4.5—
    GPT-5.40%
    Source

    Not directly comparable

Coding

  • SWE-bench Verified

    Claude Sonnet 4.577.2%
    Source
    GPT-5.4—

    Not directly comparable

  • Terminal-Bench 2.0

    Claude Sonnet 4.550.0%
    Source
    GPT-5.475.1%
    Source

    GPT-5.4 leads this result

  • LiveCodeBench Pro

    Claude Sonnet 4.5—
    GPT-5.487.5%
    Source

    Not directly comparable

  • SWE-bench Pro

    Claude Sonnet 4.5—
    GPT-5.457.7%
    Source

    Not directly comparable

  • React Native Evals

    Claude Sonnet 4.5—
    GPT-5.485.3%
    Source

    Not directly comparable

  • Vibe Code Bench

    Claude Sonnet 4.5—
    GPT-5.467.42%
    Source

    Not directly comparable

  • PostTrainBench v1.1

    Claude Sonnet 4.5—
    GPT-5.419.0%
    Source

    Not directly comparable

Reasoning

  • ARC-AGI-2

    Claude Sonnet 4.513.6%
    Source
    GPT-5.474.0%
    Source

    GPT-5.4 leads this result

  • ARC-AGI-3

    Claude Sonnet 4.5—
    GPT-5.40.2%
    Source

    Not directly comparable

  • ARC-AGI-1

    Claude Sonnet 4.5—
    GPT-5.493.67%
    Source

    Not directly comparable

Multimodal

  • MMMU-Pro

    Claude Sonnet 4.5—
    GPT-5.481.2%
    Source

    Not directly comparable

  • OfficeQA Pro

    Claude Sonnet 4.5—
    GPT-5.453.2%
    Source

    Not directly comparable

  • MMMU-Pro w/ Python

    Claude Sonnet 4.5—
    GPT-5.482.1%
    Source

    Not directly comparable

  • CharXiv

    Claude Sonnet 4.5—
    GPT-5.482.8%
    Source

    Not directly comparable

  • ERQA

    Claude Sonnet 4.5—
    GPT-5.465.4%
    Source

    Not directly comparable

  • SimpleVQA

    Claude Sonnet 4.5—
    GPT-5.461.1%
    Source

    Not directly comparable

  • ScreenSpot Pro

    Claude Sonnet 4.5—
    GPT-5.485.4%
    Source

    Not directly comparable

  • ZeroBench

    Claude Sonnet 4.5—
    GPT-5.441.0%
    Source

    Not directly comparable

  • MedXpertQA (MM)

    Claude Sonnet 4.5—
    GPT-5.477.1%
    Source

    Not directly comparable

Knowledge

  • GPQA

    Claude Sonnet 4.583.4%
    Source
    GPT-5.492.8%
    Source

    GPT-5.4 leads this result

  • HLE

    Claude Sonnet 4.5—
    GPT-5.452.1%
    Source

    Not directly comparable

  • HLE w/o tools

    Claude Sonnet 4.5—
    GPT-5.439.8%
    Source

    Not directly comparable

  • GPQA-D

    Claude Sonnet 4.5—
    GPT-5.492.8%
    Source

    Not directly comparable

  • HealthBench Hard

    Claude Sonnet 4.5—
    GPT-5.440.1%
    Source

    Not directly comparable

  • MedXpertQA (Text)

    Claude Sonnet 4.5—
    GPT-5.459.6%
    Source

    Not directly comparable

  • HealthBench Professional

    Claude Sonnet 4.5—
    GPT-5.448.1%
    Source

    Not directly comparable

Math

  • AIME 2025

    Claude Sonnet 4.587%
    Source
    GPT-5.4—

    Not directly comparable

  • FrontierMath v2 (Tiers 1-3)

    Shared source
    Claude Sonnet 4.523.860%
    GPT-5.478.596%

    GPT-5.4 leads this result

  • FrontierMath v2 (Tier 4)

    Shared source
    Claude Sonnet 4.52.439%
    GPT-5.449.000%

    GPT-5.4 leads this result

45 public results · 9 shared

Related comparisons

Compare API pricingRead the methodologyOpen the model selector

Watch Claude Sonnet 4.5 vs GPT-5.4

One weekly email when material rank, price, or benchmark evidence changes make this matchup worth revisiting.

Read a sample issue

Join 5,600+ readers.

Last updated October 17, 2026