Skip to main content

Head-to-head comparison

GPT-5.4 mini vs. Mistral Large 3

Compare published benchmark results from matching versions and cohorts, API pricing, and technical specifications. Any documented test setup differences remain visible.

shared result series
29
distinct benchmarks
1
latest retrieval
2026-09-04

Change model selection

Specifications

General specifications

Release date, availability, context window, architecture, and published parameter counts at a glance. Undisclosed details remain marked as unavailable.

Specifications for GPT-5.4 mini and Mistral Large 3
AttributeGPT-5.4 miniMistral Large 3
ProviderOpenAIMistral AI
ReleasedMar 2026Dec 2, 2025
AvailabilityActiveActive
Model typeProprietaryOpen Weights
ParametersNot published675B, 41B active
ArchitectureNot publishedMixture of Experts
Context window400,000 tokens256,000 tokens
Knowledge cutoffAugust 31, 2025Not published
Technical sourcesModelModel

Pricing

API pricing compared

The table compares official standard API rates per 1M tokens, including published cache pricing. Batch, fast, priority, regional surcharges, tool calls, and cloud platform pricing are excluded.

API pricing for GPT-5.4 mini and Mistral Large 3
Price typeGPT-5.4 miniMistral Large 3
API input$0.75$0.5
API output$4.5$1.5
Cache read$0.07Not listed
Cache write (5 min.)Not listedNot listed
Cache write (1 hr.)Not listedNot listed
Price verified07/30/2026 Source07/30/2026 Source
GPT-5.4 miniMistral Large 3
API inputUSD per 1M tokens, base tier
GPT-5.4 mini$0.75
Mistral Large 3Lower price$0.5
API outputUSD per 1M tokens, base tier
GPT-5.4 mini$4.5
Mistral Large 3Lower price$1.5

Key to the marks:better valuebetter, but not conclusivetie

Pricing by provider

Direct developer pricing and provider endpoint tariffs routed through OpenRouter are listed separately. Prices are in USD per 1M tokens. Context-dependent direct tiers plus regional, Flex, Priority, and other OpenRouter tariffs remain individually identifiable.

Provider pricing for GPT-5.4 mini and Mistral Large 3
ModelProviderPurchase route and tariffInput per 1M tokensOutput per 1M tokensSource
GPT-5.4 miniOpenAIDirect from developerStandard API$0.75$4.5Source07/30/2026
AzureVia OpenRouterazure$0.75$4.5Source09/07/2026
AzureVia OpenRouterazure/us$0.825$4.95Source09/07/2026
OpenAIVia OpenRouteropenai$0.75$4.5Source09/07/2026
OpenAIVia OpenRouteropenai/fast$1.5$9Source09/07/2026
OpenAIVia OpenRouteropenai/flexLowest listed price: $0.375Lowest listed price: $2.25Source09/07/2026
Mistral Large 3Mistral AIDirect from developerStandard APILowest listed price: $0.5Lowest listed price: $1.5Source07/30/2026
MistralVia OpenRoutermistralLowest listed price: $0.5Lowest listed price: $1.5Source09/07/2026
MistralVia OpenRoutermistral/eu$0.55$1.65Source09/07/2026
MistralVia OpenRoutermistral/zdrLowest listed price: $0.5Lowest listed price: $1.5Source09/07/2026

Performance

Shared benchmarks

Only published results with the same benchmark version, task, metric, and comparison cohort are paired. Different reasoning levels, agents, harnesses, or output limits appear directly in the table.

Capability profile from matched benchmarks

Each axis averages directly matched benchmark families. A value of 100 means the stronger value within this pair, not a universal quality score.

Not enough directly matched categories for a reliable radar

Only 0 of at least 5 categories have comparable, direct model measurements. The benchmark tables below show the available evidence without filling missing axes.

GPT-5.4 miniMistral Large 3
LMArena Text, Style ControlOverall. Higher is better. Test setup not documented as identical
GPT-5.4 miniHigher measured value1,448.118
Mistral Large 31,413.556
LMArena Text, Style ControlBusiness, management and finance. Higher is better. Test setup not documented as identical
GPT-5.4 miniHigher measured value1,459.508
Mistral Large 31,412.721
LMArena Text, Style ControlChinese. Higher is better. Test setup not documented as identical
GPT-5.4 miniHigher measured value1,477.451
Mistral Large 31,431.759
LMArena Text, Style ControlCoding. Higher is better. Test setup not documented as identical
GPT-5.4 miniHigher measured value1,495.819
Mistral Large 31,468.587
LMArena Text, Style ControlCreative writing. Higher is better. Test setup not documented as identical
GPT-5.4 miniHigher measured value1,403.981
Mistral Large 31,375.017
LMArena Text, Style ControlEnglish. Higher is better. Test setup not documented as identical
GPT-5.4 miniHigher measured value1,450.856
Mistral Large 31,426.818
LMArena Text, Style ControlEntertainment, sports and media. Higher is better. Test setup not documented as identical
GPT-5.4 miniHigher measured value1,410.7
Mistral Large 31,373.228
LMArena Text, Style ControlExcluding ties. Higher is better. Test setup not documented as identical
GPT-5.4 miniHigher measured value1,449.951
Mistral Large 31,400.876
LMArena Text, Style ControlExpert prompts. Higher is better. Test setup not documented as identical
GPT-5.4 miniHigher measured value1,480.528
Mistral Large 31,423.349
LMArena Text, Style ControlFrench. Higher is better. Test setup not documented as identical
GPT-5.4 miniHigher measured value1,474.797
Mistral Large 31,434.923
LMArena Text, Style ControlGerman. Higher is better. Test setup not documented as identical
GPT-5.4 miniHigher measured value1,442.29
Mistral Large 31,404.512
LMArena Text, Style ControlHard prompts. Higher is better. Test setup not documented as identical
GPT-5.4 miniHigher measured value1,469.848
Mistral Large 31,431.026

Key to the marks:better valuebetter, but not conclusivetie

Benchmark scores for GPT-5.4 mini and Mistral Large 3
Benchmark and taskGPT-5.4 miniMistral Large 3Source
LMArena Text, Style ControlOverall
Test details
Test setup not documented as identicalNo direct winnerVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterGPT-5.4 mini: Snapshot: gpt-5.4-mini-high, Reasoning: high, Harness: Arena text, style control
Mistral Large 3: Snapshot: mistral-large-3, Harness: Arena text, style controlGPT-5.4 mini: Style-controlled Arena rating from 59,472 anonymous pairwise votes.Mistral Large 3: Style-controlled Arena rating from 65,336 anonymous pairwise votes.
Higher measured value1,448.118
95% confidence interval: ±3.851The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
1,413.556
95% confidence interval: ±3.08The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.		LMArena
Source details
Data as of: 09/01/2026
LMArena Text, Style ControlBusiness, management and finance
Test details
Test setup not documented as identicalNo direct winnerVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterGPT-5.4 mini: Snapshot: gpt-5.4-mini-high, Reasoning: high, Harness: Arena text, style control
Mistral Large 3: Snapshot: mistral-large-3, Harness: Arena text, style controlGPT-5.4 mini: Style-controlled Arena rating from 12,113 anonymous pairwise votes.Mistral Large 3: Style-controlled Arena rating from 12,392 anonymous pairwise votes.
Higher measured value1,459.508
95% confidence interval: ±6.653The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
1,412.721
95% confidence interval: ±5.917The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.		LMArena
Source details
Data as of: 09/01/2026
LMArena Text, Style ControlChinese
Test details
Test setup not documented as identicalNo direct winnerVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterGPT-5.4 mini: Snapshot: gpt-5.4-mini-high, Reasoning: high, Harness: Arena text, style control
Mistral Large 3: Snapshot: mistral-large-3, Harness: Arena text, style controlGPT-5.4 mini: Style-controlled Arena rating from 3,298 anonymous pairwise votes.Mistral Large 3: Style-controlled Arena rating from 4,483 anonymous pairwise votes.
Higher measured value1,477.451
95% confidence interval: ±11.125The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
1,431.759
95% confidence interval: ±9.282The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.		LMArena
Source details
Data as of: 09/01/2026
LMArena Text, Style ControlCoding
Test details
Test setup not documented as identicalNo direct winnerVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterGPT-5.4 mini: Snapshot: gpt-5.4-mini-high, Reasoning: high, Harness: Arena text, style control
Mistral Large 3: Snapshot: mistral-large-3, Harness: Arena text, style controlGPT-5.4 mini: Style-controlled Arena rating from 16,204 anonymous pairwise votes.Mistral Large 3: Style-controlled Arena rating from 16,301 anonymous pairwise votes.
Higher measured value1,495.819
95% confidence interval: ±6.028The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
1,468.587
95% confidence interval: ±5.241The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.		LMArena
Source details
Data as of: 09/01/2026
LMArena Text, Style ControlCreative writing
Test details
Test setup not documented as identicalNo direct winnerVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterGPT-5.4 mini: Snapshot: gpt-5.4-mini-high, Reasoning: high, Harness: Arena text, style control
Mistral Large 3: Snapshot: mistral-large-3, Harness: Arena text, style controlGPT-5.4 mini: Style-controlled Arena rating from 9,837 anonymous pairwise votes.Mistral Large 3: Style-controlled Arena rating from 10,954 anonymous pairwise votes.
Higher measured value1,403.981
95% confidence interval: ±7.287The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
1,375.017
95% confidence interval: ±6.318The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.		LMArena
Source details
Data as of: 09/01/2026
LMArena Text, Style ControlEnglish
Test details
Test setup not documented as identicalNo direct winnerVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterGPT-5.4 mini: Snapshot: gpt-5.4-mini-high, Reasoning: high, Harness: Arena text, style control
Mistral Large 3: Snapshot: mistral-large-3, Harness: Arena text, style controlGPT-5.4 mini: Style-controlled Arena rating from 27,483 anonymous pairwise votes.Mistral Large 3: Style-controlled Arena rating from 27,473 anonymous pairwise votes.
Higher measured value1,450.856
95% confidence interval: ±5.045The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
1,426.818
95% confidence interval: ±4.282The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.		LMArena
Source details
Data as of: 09/01/2026
LMArena Text, Style ControlEntertainment, sports and media
Test details
Test setup not documented as identicalNo direct winnerVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterGPT-5.4 mini: Snapshot: gpt-5.4-mini-high, Reasoning: high, Harness: Arena text, style control
Mistral Large 3: Snapshot: mistral-large-3, Harness: Arena text, style controlGPT-5.4 mini: Style-controlled Arena rating from 12,709 anonymous pairwise votes.Mistral Large 3: Style-controlled Arena rating from 13,723 anonymous pairwise votes.
Higher measured value1,410.7
95% confidence interval: ±6.674The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
1,373.228
95% confidence interval: ±5.71The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.		LMArena
Source details
Data as of: 09/01/2026
LMArena Text, Style ControlExcluding ties
Test details
Test setup not documented as identicalNo direct winnerVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterGPT-5.4 mini: Snapshot: gpt-5.4-mini-high, Reasoning: high, Harness: Arena text, style control
Mistral Large 3: Snapshot: mistral-large-3, Harness: Arena text, style controlGPT-5.4 mini: Style-controlled Arena rating from 45,611 anonymous pairwise votes.Mistral Large 3: Style-controlled Arena rating from 47,430 anonymous pairwise votes.
Higher measured value1,449.951
95% confidence interval: ±5.017The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
1,400.876
95% confidence interval: ±4.294The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.		LMArena
Source details
Data as of: 09/01/2026
LMArena Text, Style ControlExpert prompts
Test details
Test setup not documented as identicalNo direct winnerVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterGPT-5.4 mini: Snapshot: gpt-5.4-mini-high, Reasoning: high, Harness: Arena text, style control
Mistral Large 3: Snapshot: mistral-large-3, Harness: Arena text, style controlGPT-5.4 mini: Style-controlled Arena rating from 5,779 anonymous pairwise votes.Mistral Large 3: Style-controlled Arena rating from 5,705 anonymous pairwise votes.
Higher measured value1,480.528
95% confidence interval: ±8.64The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
1,423.349
95% confidence interval: ±8.346The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.		LMArena
Source details
Data as of: 09/01/2026
LMArena Text, Style ControlFrench
Test details
Test setup not documented as identicalNo direct winnerVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterGPT-5.4 mini: Snapshot: gpt-5.4-mini-high, Reasoning: high, Harness: Arena text, style control
Mistral Large 3: Snapshot: mistral-large-3, Harness: Arena text, style controlGPT-5.4 mini: Style-controlled Arena rating from 2,126 anonymous pairwise votes.Mistral Large 3: Style-controlled Arena rating from 1,858 anonymous pairwise votes.
Higher measured value1,474.797
95% confidence interval: ±15.191The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
1,434.923
95% confidence interval: ±15.523The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.		LMArena
Source details
Data as of: 09/01/2026
LMArena Text, Style ControlGerman
Test details
Test setup not documented as identicalNo direct winnerVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterGPT-5.4 mini: Snapshot: gpt-5.4-mini-high, Reasoning: high, Harness: Arena text, style control
Mistral Large 3: Snapshot: mistral-large-3, Harness: Arena text, style controlGPT-5.4 mini: Style-controlled Arena rating from 898 anonymous pairwise votes.Mistral Large 3: Style-controlled Arena rating from 1,080 anonymous pairwise votes.
Higher measured value1,442.29
95% confidence interval: ±20.853The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
1,404.512
95% confidence interval: ±19.03The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.		LMArena
Source details
Data as of: 09/01/2026
LMArena Text, Style ControlHard prompts
Test details
Test setup not documented as identicalNo direct winnerVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterGPT-5.4 mini: Snapshot: gpt-5.4-mini-high, Reasoning: high, Harness: Arena text, style control
Mistral Large 3: Snapshot: mistral-large-3, Harness: Arena text, style controlGPT-5.4 mini: Style-controlled Arena rating from 38,922 anonymous pairwise votes.Mistral Large 3: Style-controlled Arena rating from 39,237 anonymous pairwise votes.
Higher measured value1,469.848
95% confidence interval: ±4.604The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
1,431.026
95% confidence interval: ±3.848The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.		LMArena
Source details
Data as of: 09/01/2026
LMArena Text, Style ControlHard prompts, English
Test details
Test setup not documented as identicalNo direct winnerVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterGPT-5.4 mini: Snapshot: gpt-5.4-mini-high, Reasoning: high, Harness: Arena text, style control
Mistral Large 3: Snapshot: mistral-large-3, Harness: Arena text, style controlGPT-5.4 mini: Style-controlled Arena rating from 18,738 anonymous pairwise votes.Mistral Large 3: Style-controlled Arena rating from 16,526 anonymous pairwise votes.
Higher measured value1,466.755
95% confidence interval: ±5.803The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
1,438.338
95% confidence interval: ±5.247The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.		LMArena
Source details
Data as of: 09/01/2026
LMArena Text, Style ControlInstruction following
Test details
Test setup not documented as identicalNo direct winnerVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterGPT-5.4 mini: Snapshot: gpt-5.4-mini-high, Reasoning: high, Harness: Arena text, style control
Mistral Large 3: Snapshot: mistral-large-3, Harness: Arena text, style controlGPT-5.4 mini: Style-controlled Arena rating from 20,183 anonymous pairwise votes.Mistral Large 3: Style-controlled Arena rating from 20,253 anonymous pairwise votes.
Higher measured value1,434.333
95% confidence interval: ±5.574The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
1,399.916
95% confidence interval: ±4.811The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.		LMArena
Source details
Data as of: 09/01/2026
LMArena Text, Style ControlJapanese
Test details
Test setup not documented as identicalNo direct winnerVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterGPT-5.4 mini: Snapshot: gpt-5.4-mini-high, Reasoning: high, Harness: Arena text, style control
Mistral Large 3: Snapshot: mistral-large-3, Harness: Arena text, style controlGPT-5.4 mini: Style-controlled Arena rating from 588 anonymous pairwise votes.Mistral Large 3: Style-controlled Arena rating from 661 anonymous pairwise votes.
Higher measured value1,409.204
95% confidence interval: ±26.147The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
1,382.813
95% confidence interval: ±24.568The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.		LMArena
Source details
Data as of: 09/01/2026
LMArena Text, Style ControlKorean
Test details
Test setup not documented as identicalNo direct winnerVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterGPT-5.4 mini: Snapshot: gpt-5.4-mini-high, Reasoning: high, Harness: Arena text, style control
Mistral Large 3: Snapshot: mistral-large-3, Harness: Arena text, style controlGPT-5.4 mini: Style-controlled Arena rating from 962 anonymous pairwise votes.Mistral Large 3: Style-controlled Arena rating from 1,118 anonymous pairwise votes.
Higher measured value1,397.676
95% confidence interval: ±20.641The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
1,355.204
95% confidence interval: ±19.178The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.		LMArena
Source details
Data as of: 09/01/2026
LMArena Text, Style ControlLegal and government
Test details
Test setup not documented as identicalNo direct winnerVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterGPT-5.4 mini: Snapshot: gpt-5.4-mini-high, Reasoning: high, Harness: Arena text, style control
Mistral Large 3: Snapshot: mistral-large-3, Harness: Arena text, style controlGPT-5.4 mini: Style-controlled Arena rating from 4,743 anonymous pairwise votes.Mistral Large 3: Style-controlled Arena rating from 4,880 anonymous pairwise votes.
Higher measured value1,456.149
95% confidence interval: ±9.594The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
1,420.347
95% confidence interval: ±8.946The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.		LMArena
Source details
Data as of: 09/01/2026
LMArena Text, Style ControlLife, physical and social science
Test details
Test setup not documented as identicalNo direct winnerVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterGPT-5.4 mini: Snapshot: gpt-5.4-mini-high, Reasoning: high, Harness: Arena text, style control
Mistral Large 3: Snapshot: mistral-large-3, Harness: Arena text, style controlGPT-5.4 mini: Style-controlled Arena rating from 9,789 anonymous pairwise votes.Mistral Large 3: Style-controlled Arena rating from 10,595 anonymous pairwise votes.
Higher measured value1,466.377
95% confidence interval: ±7.074The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
1,427.433
95% confidence interval: ±6.173The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.		LMArena
Source details
Data as of: 09/01/2026
LMArena Text, Style ControlLonger queries
Test details
Test setup not documented as identicalNo direct winnerVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterGPT-5.4 mini: Snapshot: gpt-5.4-mini-high, Reasoning: high, Harness: Arena text, style control
Mistral Large 3: Snapshot: mistral-large-3, Harness: Arena text, style controlGPT-5.4 mini: Style-controlled Arena rating from 25,449 anonymous pairwise votes.Mistral Large 3: Style-controlled Arena rating from 23,193 anonymous pairwise votes.
Higher measured value1,447.807
95% confidence interval: ±5.419The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
1,416.784
95% confidence interval: ±4.746The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.		LMArena
Source details
Data as of: 09/01/2026
LMArena Text, Style ControlMath
Test details
Test setup not documented as identicalNo direct winnerVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterGPT-5.4 mini: Snapshot: gpt-5.4-mini-high, Reasoning: high, Harness: Arena text, style control
Mistral Large 3: Snapshot: mistral-large-3, Harness: Arena text, style controlGPT-5.4 mini: Style-controlled Arena rating from 3,165 anonymous pairwise votes.Mistral Large 3: Style-controlled Arena rating from 3,755 anonymous pairwise votes.
Higher measured value1,440.49
95% confidence interval: ±11.097The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
1,402.864
95% confidence interval: ±9.902The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.		LMArena
Source details
Data as of: 09/01/2026
LMArena Text, Style ControlMathematical professions
Test details
Test setup not documented as identicalNo direct winnerVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterGPT-5.4 mini: Snapshot: gpt-5.4-mini-high, Reasoning: high, Harness: Arena text, style control
Mistral Large 3: Snapshot: mistral-large-3, Harness: Arena text, style controlGPT-5.4 mini: Style-controlled Arena rating from 3,139 anonymous pairwise votes.Mistral Large 3: Style-controlled Arena rating from 3,175 anonymous pairwise votes.
Higher measured value1,453.017
95% confidence interval: ±11.39The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
1,417.813
95% confidence interval: ±10.98The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.		LMArena
Source details
Data as of: 09/01/2026
LMArena Text, Style ControlMedicine and healthcare
Test details
Test setup not documented as identicalNo direct winnerVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterGPT-5.4 mini: Snapshot: gpt-5.4-mini-high, Reasoning: high, Harness: Arena text, style control
Mistral Large 3: Snapshot: mistral-large-3, Harness: Arena text, style controlGPT-5.4 mini: Style-controlled Arena rating from 4,429 anonymous pairwise votes.Mistral Large 3: Style-controlled Arena rating from 4,505 anonymous pairwise votes.
Higher measured value1,452.045
95% confidence interval: ±9.996The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
1,431.127
95% confidence interval: ±9.388The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.		LMArena
Source details
Data as of: 09/01/2026
LMArena Text, Style ControlMulti-turn
Test details
Test setup not documented as identicalNo direct winnerVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterGPT-5.4 mini: Snapshot: gpt-5.4-mini-high, Reasoning: high, Harness: Arena text, style control
Mistral Large 3: Snapshot: mistral-large-3, Harness: Arena text, style controlGPT-5.4 mini: Style-controlled Arena rating from 11,173 anonymous pairwise votes.Mistral Large 3: Style-controlled Arena rating from 11,201 anonymous pairwise votes.
Higher measured value1,468.338
95% confidence interval: ±6.838The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
1,420.712
95% confidence interval: ±6.145The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.		LMArena
Source details
Data as of: 09/01/2026
LMArena Text, Style ControlNon-English
Test details
Test setup not documented as identicalNo direct winnerVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterGPT-5.4 mini: Snapshot: gpt-5.4-mini-high, Reasoning: high, Harness: Arena text, style control
Mistral Large 3: Snapshot: mistral-large-3, Harness: Arena text, style controlGPT-5.4 mini: Style-controlled Arena rating from 31,988 anonymous pairwise votes.Mistral Large 3: Style-controlled Arena rating from 37,859 anonymous pairwise votes.
Higher measured value1,438.783
95% confidence interval: ±4.78The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
1,397.941
95% confidence interval: ±3.805The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.		LMArena
Source details
Data as of: 09/01/2026
LMArena Text, Style ControlPolish
Test details
Test setup not documented as identicalNo direct winnerVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterGPT-5.4 mini: Snapshot: gpt-5.4-mini-high, Reasoning: high, Harness: Arena text, style control
Mistral Large 3: Snapshot: mistral-large-3, Harness: Arena text, style controlGPT-5.4 mini: Style-controlled Arena rating from 1,164 anonymous pairwise votes.Mistral Large 3: Style-controlled Arena rating from 1,582 anonymous pairwise votes.
Higher measured value1,460.824
95% confidence interval: ±18.171The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
1,415.726
95% confidence interval: ±15.021The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.		LMArena
Source details
Data as of: 09/01/2026
LMArena Text, Style ControlRussian
Test details
Test setup not documented as identicalNo direct winnerVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterGPT-5.4 mini: Snapshot: gpt-5.4-mini-high, Reasoning: high, Harness: Arena text, style control
Mistral Large 3: Snapshot: mistral-large-3, Harness: Arena text, style controlGPT-5.4 mini: Style-controlled Arena rating from 6,279 anonymous pairwise votes.Mistral Large 3: Style-controlled Arena rating from 7,429 anonymous pairwise votes.
Higher measured value1,454.249
95% confidence interval: ±8.318The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
1,409.803
95% confidence interval: ±7.118The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.		LMArena
Source details
Data as of: 09/01/2026
LMArena Text, Style ControlSoftware and IT services
Test details
Test setup not documented as identicalNo direct winnerVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterGPT-5.4 mini: Snapshot: gpt-5.4-mini-high, Reasoning: high, Harness: Arena text, style control
Mistral Large 3: Snapshot: mistral-large-3, Harness: Arena text, style controlGPT-5.4 mini: Style-controlled Arena rating from 23,530 anonymous pairwise votes.Mistral Large 3: Style-controlled Arena rating from 24,350 anonymous pairwise votes.
Higher measured value1,485.976
95% confidence interval: ±5.317The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
1,456.163
95% confidence interval: ±4.498The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.		LMArena
Source details
Data as of: 09/01/2026
LMArena Text, Style ControlSpanish
Test details
Test setup not documented as identicalNo direct winnerVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterGPT-5.4 mini: Snapshot: gpt-5.4-mini-high, Reasoning: high, Harness: Arena text, style control
Mistral Large 3: Snapshot: mistral-large-3, Harness: Arena text, style controlGPT-5.4 mini: Style-controlled Arena rating from 1,860 anonymous pairwise votes.Mistral Large 3: Style-controlled Arena rating from 1,998 anonymous pairwise votes.
Higher measured value1,436.164
95% confidence interval: ±15.348The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
1,406.852
95% confidence interval: ±14.343The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.		LMArena
Source details
Data as of: 09/01/2026
LMArena Text, Style ControlWriting, literature and language
Test details
Test setup not documented as identicalNo direct winnerVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterGPT-5.4 mini: Snapshot: gpt-5.4-mini-high, Reasoning: high, Harness: Arena text, style control
Mistral Large 3: Snapshot: mistral-large-3, Harness: Arena text, style controlGPT-5.4 mini: Style-controlled Arena rating from 14,562 anonymous pairwise votes.Mistral Large 3: Style-controlled Arena rating from 15,693 anonymous pairwise votes.
Higher measured value1,425.044
95% confidence interval: ±6.215The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.
1,390.534
95% confidence interval: ±5.355The Arena score measures human preference rather than a fixed capability test. The confidence interval and model configuration are part of the result.		LMArena
Source details
Data as of: 09/01/2026

Key to the marks:better valuebetter, but not conclusivetie

How to read this comparison

Winning one benchmark is not an overall verdict

Your workload, budget, and required context length matter most. A coding benchmark says little about visual understanding or agent performance.

The standard price table uses direct developer rates. The provider comparison labels OpenRouter endpoints, regions, and special tariffs separately.

Every result links to its measurement source and retrieval date. Labels distinguish vendor reports, official benchmark leaderboards, and independent evaluations. Disputed or archived results are excluded.

More matchups

Related LLM comparisons

Compare GPT-5.4 mini and Mistral Large 3 with other leading models using the same data and benchmark logic.

View the complete LLM comparison