BenchLeader
Menu
Best LLMsChoose a modelCoding plansLeaderboardsModelsBenchmarksProvidersReleases7

Gemini 3.1 Pro vs GPT-5.4

Full profiles: Gemini 3.1 Pro (Google) and GPT-5.4 (OpenAI): every benchmark, price and rank.

Updated · Data from 15 sources

As of 6 Nov 2026, Gemini 3.1 Pro and GPT-5.4 are neck and neck on quality (60.5 vs 63.2 on the BenchLeader Index, inside the ±4.9 margin), and Gemini 3.1 Pro is 1.3× cheaper at $4.50 vs $5.63 per 1M tokens blended. Choose Gemini 3.1 Pro for multimodal (65 vs 61) and fast first answers (18.2 s vs 147.3 s), where it also streams faster; choose GPT-5.4 for coding (65 vs 57), agents & tools (63 vs 57), and instruction following (72 vs 67).

Key facts (as of 6 Nov 2026)

  • Gemini 3.1 Pro scores 60.5 on the BenchLeader Index and GPT-5.4 scores 63.2 (BenchLeader, 6 Nov 2026).
  • Gemini 3.1 Pro costs $4.50 per 1M tokens blended (list price, 6 Nov 2026).
  • GPT-5.4 costs $5.63 per 1M tokens blended (list price, 6 Nov 2026).
  • Gemini 3.1 Pro streams faster than GPT-5.4 (Artificial Analysis, 7-day median, 6 Nov 2026).
  • Of the 35 benchmarks both models have, GPT-5.4 leads 21 (BenchLeader, 6 Nov 2026).
Gemini 3.1 Pro benchmarks Google GPT-5.4 benchmarks xhigh OpenAIAdd a third model

Which to pick for…

The pick's numbers come first. Category scores on the BenchLeader scale (50 is the average model, 15 points one standard deviation). High confidence means a lead of half a standard deviation or more; under one point is called either.

Use casePickWhyConfidence
CodingGPT-5.465 vs 57 category score; WeirdML 77.7% vs 72.1%High
Agents & tool useGPT-5.463 vs 57 category score; τ³-Banking (AA) 39.6% vs 21.4%Medium
MathsGPT-5.465 vs 61 category score; FrontierMath Tiers 1–3 78.6% vs 59.6%Medium
ReasoningEither66 vs 66 category score; GPQA Diamond 94.1% vs 93.3%Either
KnowledgeGPT-5.459 vs 56 category score; EBR-bench 25.4% vs 14.3%Medium
Long documentsEither66 vs 66 category score; AA-LCR v1.1 82.0% vs 82.0%Either
Images and other mediaGemini 3.1 Pro65 vs 61 category score; MMMU-Pro 82.4% vs 78.4%Medium
Following instructionsGPT-5.472 vs 67 category score; IFBench 74.0% vs 77.1%Medium
Chat quality (human preference)–Only Gemini 3.1 Pro has an Arena rating (58 category score)Not comparable
Lowest costGemini 3.1 Pro$4.50 vs $5.63 per 1M blended; agentic coding session $0.168 vs $0.210–
Fastest first answerGemini 3.1 Pro18.2 s (Gemini 3.1 Pro) vs 147.3 s (GPT-5.4 (xhigh)) to the first answer token–
High volume on a budgetNeither flagshipA smaller model is usually the better fit: Cheapest good LLMs, Fastest LLMs–

Key differences

  • Gemini 3.1 Pro and GPT-5.4 (xhigh) are neck and neck on quality (60.5 vs 63.2, within the ±4.9 margin).
  • Gemini 3.1 Pro is stronger in human preference, multimodal, reasoning.
  • GPT-5.4 (xhigh) is stronger in agents & tools, coding, instruction following, knowledge, maths.
  • Gemini 3.1 Pro is 1.3× cheaper ($4.50 vs $5.63 per 1M blended).
  • Gemini 3.1 Pro streams 1.9× faster (122 vs 64 tokens per second).

At a glance

SpecGemini 3.1 ProGPT-5.4
DeveloperGoogleOpenAI
Released19 Feb 20265 Mar 2026
Context window (tokens)1,048,5761,050,000
Compared atGemini 3.1 ProGPT-5.4 (xhigh)
Open weightsNoNo
Price in / out / cached input per 1M$2.00 / $12.00 / $0.200$2.50 / $15.00 / $0.250

Category by category

Category scores on the BenchLeader scale: 50 is the average evaluated model, 15 points is one standard deviation. A gap under one point is called level.

CategoryGemini 3.1 ProGPT-5.4 (xhigh)Leader
Reasoning66.366.0level
Coding56.764.5GPT-5.4
Agents & tools57.163.0GPT-5.4
Maths60.765.1GPT-5.4
Knowledge55.858.7GPT-5.4
Instruction following66.972.3GPT-5.4
Human preference58.4–not comparable
Multimodal64.560.9Gemini 3.1 Pro
Long context65.965.9level

On one chart

Each domain on the index's own scale, where 50 is an average model. Click a domain to see the benchmarks behind it. Pick up to 5 models from the list.

2550 avg75ReasoningCodingAgents & toolsMathsKnowledgeInstructionfollowingLong contextMultimodalGemini 3.1 Pro, Reasoning: 66.3Gemini 3.1 Pro, Coding: 56.7Gemini 3.1 Pro, Agents & tools: 57.1Gemini 3.1 Pro, Maths: 60.7Gemini 3.1 Pro, Knowledge: 55.8Gemini 3.1 Pro, Instruction following: 66.9Gemini 3.1 Pro, Long context: 65.9Gemini 3.1 Pro, Multimodal: 64.5GPT-5.4, Reasoning: 66.0GPT-5.4, Coding: 64.5GPT-5.4, Agents & tools: 63.0GPT-5.4, Maths: 65.1GPT-5.4, Knowledge: 58.7GPT-5.4, Instruction following: 72.3GPT-5.4, Long context: 65.9GPT-5.4, Multimodal: 60.9

Not drawn, measured for fewer than two of these models: Human preference.

ModelIndex

Where sources disagree

Across the 35 benchmarks both have, Gemini 3.1 Pro leads 13, GPT-5.4 leads 21 and 1 is a tie. Every overall index that rates both puts Gemini 3.1 Pro ahead.

Overall indexGemini 3.1 ProGPT-5.4 (xhigh)Leader
BenchLeader Index60.563.2neck and neck
Arena Text14871475†Gemini 3.1 Pro

Each publisher runs its own tasks, harnesses and effort settings, so their overall verdicts can differ; BenchLeader reconciles them on one scale. † measured at another effort setting of the same model (the publisher did not test the one compared here).

Price, speed and latency

List prices blended at three input tokens per output token. Per-task prices use the token counts on the methodology page. Cost, tokens and time per benchmark task are measured: Artificial Analysis’ averages over its whole Intelligence Index run.

MeasureGemini 3.1 ProGPT-5.4 (xhigh)
Blended price per 1M tokens$4.50$5.63
Input / output per 1M$2.00 / $12.00$2.50 / $15.00
Cached input per 1M$0.200$0.250
Price per task: chat reply$0.0044$0.0055
Price per task: summarize a 30-page report$0.031$0.039
Price per task: code edit$0.030$0.037
Price per task: agentic coding session$0.168$0.210
Price per task: structured extraction$0.0064$0.0080
Agentic coding session, input mostly cached$0.087$0.109
Output speed122 tok/s64 tok/s
Time to first answer18.2 s147.3 s
Time to a complete answer43.8 s237.8 s

Each model is compared at its best-ranked reasoning-effort setting. Gemini 3.1 Pro scores 60.5 and takes 18 s to its first answer token; Gemini 3.1 Pro (high) scores 60.1 in 18 s. GPT-5.4 (xhigh) scores 63.2 and takes 147 s to its first answer token; GPT-5.4 (low) scores 58.9 in 11 s, GPT-5.4 (high) scores 58.8 in 2 s. Higher effort means more thinking before the answer: better scores, longer waits and more tokens billed. How reasoning effort changes scores and cost.

Every benchmark

154 benchmarks, grouped by category; each name links to the board and its leaders. † measured at another effort setting of the same model, shown for reference and not counted as a win.

MetricPublisherGemini 3.1 ProGPT-5.4 (xhigh)
BenchLeader IndexBenchLeader60.563.2
Agents & tools scoreBenchLeader57.163.0
Coding scoreBenchLeader56.764.5
Human preference scoreBenchLeader58.4–
Instruction following scoreBenchLeader66.972.3
Knowledge scoreBenchLeader55.858.7
Long context scoreBenchLeader65.965.9
Maths scoreBenchLeader60.765.1
Multimodal scoreBenchLeader64.560.9
Reasoning scoreBenchLeader66.366.0
Blended price $/MBenchLeader$4.50$5.63
Output speedBenchLeader122 tok/s64 tok/s
Time to first answerBenchLeader18.2 s147.3 s
Context windowBenchLeader1,048,576 tokens1,050,000 tokens
Reasoning
GPQA DiamondEpoch AI Benchmarking Hub94.1%93.3%
Humanity's Last ExamScale AI / CAIS46.4%36.2%
SimpleBenchSimpleBench79.6%–
Arena Hard PromptsArena (LMArena)15081487†
LiveBench 1.0 ReasoningLiveBench84.0%†88.1%
GPQA Diamond (AA)Artificial Analysis94.1%92.0%
Humanity's Last Exam (AA)Artificial Analysis47.0%43.7%
HLE-Diamond (AA)Artificial Analysis31.4%–
GPQA Diamond (Vals)Vals AI95.5%†91.7%
EnigmaEvalScale AI SEAL36.8%†–
Kagi LLM BenchmarkKagi LLM Benchmark–63.8%†
ARC-AGI-1ARC Prize98.0%93.7%
ARC-AGI-2ARC Prize77.1%74.0%
ARC-AGI-3ARC Prize0.4%0.2%†
CritPtArtificial Analysis17.7%23.4%
MysteryMechanismVals AI20.7%†–
Chess PuzzlesEpoch AI Benchmarking Hub51.0%†48.0%
Mystery Game PuzzlesEpoch AI Benchmarking Hub34.0%†37.0%
CL-benchEpoch AI Benchmarking Hub20.8%27.9%
CL-bench LifeEpoch AI Benchmarking Hub16.9%21.7%
LMCAEpoch AI Benchmarking Hub53.8%†52.0%
DTBenchEpoch AI Benchmarking Hub95.5%†94.4%
ForecastBenchForecastBench64.3%63.1%†
Coding
SWE-bench Verified (Epoch)Epoch AI Benchmarking Hub–76.9%†
SciCodeSciCode58.7%56.6%
Show all 154 benchmarks
MetricPublisherGemini 3.1 ProGPT-5.4 (xhigh)
Coding
WeirdMLWeirdML72.1%77.7%
GSO-BenchGSO-Bench22.6%31.4%
Arena CodingArena (LMArena)15211511†
Arena WebDevArena (LMArena)14471401†
LiveBench 1.0 CodingLiveBench76.5%†77.5%
SciCode (AA)Artificial Analysis58.7%–
LiveCodeBenchVals AI88.5%†84.1%
IOIVals AI51.8%†–
SWE-bench (Vals)Vals AI78.8%†78.2%
SWE-Bench ProScale AI SEAL46.1%†59.1%
Code MigrationVals AI17.3%†35.0%
ProgramBenchVals AI0.0%†0.0%
Vibe Code Bench 1-100Vals AI6.7%†–
Vibe Code Bench v1.1Vals AI32.0%†67.4%
SWE-Bench Pro (private)Scale AI SEAL46.1%†59.1%
SWE Atlas: Codebase QnAScale AI SEAL13.5%40.8%
SWE Atlas: RefactoringScale AI SEAL33.8%44.3%
SWE Atlas: Test WritingScale AI SEAL29.8%44.4%
MirrorCodeEpoch AI Benchmarking Hub8.9%†15.6%†
DeepSWE v1.1Epoch AI Benchmarking Hub11.7%51.8%
ALE-BenchEpoch AI Benchmarking Hub1160.61607†
AlgoTuneAlgoTune2.01.9†
Terminal-Bench 4.0 (AA)Artificial Analysis4.0%–
Terminal-Bench 2.1 (AA)Artificial Analysis73.8%78.3%
GameBench 2 OverallDuelLab GameBench 2–53.4†
GameBench 2 by reasoning settingDuelLab GameBench 2–61.8
Agents & tools
Terminal-BenchTerminal-Bench80.2%81.8%†
APEX-AgentsMercor35.1%52.4%†
Arena AgentArena (LMArena)-4.4–
LiveBench Agentic CodingLiveBench44.1%†53.8%
Terminal-Bench HardArtificial Analysis53.8%57.6%
τ²-Bench Telecom (AA)Artificial Analysis95.6%87.1%
Terminal-Bench 2.1 (Vals)Vals AI70.8%†–
MCP AtlasScale AI SEAL78.2%†70.6%
HiL-BenchScale AI SEAL35.3%9.7%†
τ³-bench Bankingτ²-bench26.0%†39.4%
GDPval-AA v2.1, normalised scoreArtificial Analysis14.6%–
GDPval-AA v2.1Artificial Analysis799–
τ³-Banking (AA)Artificial Analysis21.4%39.6%
ITBench SRE (AA)Artificial Analysis30.3%–
Analyst Agent (AA)Artificial Analysis41.3%–
APEX-Agents (AA)Artificial Analysis32.0%33.3%
Finance Agent v2Vals AI43.0%†–
Harvey's Legal Agent Benchmark v1Vals AI0.0%†0.0%
Legal Research BenchVals AI20.7%†–
SkillsBenchVals AI–51.7%
SREBenchVals AI0.0%†–
Tax Agent BenchVals AI9.3%†–
Terminal-Bench 2.0 (Vals)Vals AI67.4%†58.4%†
Terminal-Bench 4.0 (Vals)Vals AI2.5%†–
Terminal-Bench ScienceVals AI1.4%†–
BALROGBALROG57.0%–
DeepResearch BenchDeepResearch Bench47.8%†35.1%†
PostTrainBenchEpoch AI Benchmarking Hub22.0%19.0%†
ExploitBenchEpoch AI Benchmarking Hub26.1%–
METR Time HorizonsMETR77.0%74.3%
GBAEvalEpoch AI Benchmarking Hub0.8%45.1%†
Vending-Bench 2Andon Labs911.26144.2†
AutomationBenchArtificial Analysis35.4%–
GDP.pdf (AA)Artificial Analysis17.8%–
Harvey LAB-AA v1.1Harvey0.0%–
EnterpriseOps-GymArtificial Analysis42.2%–
AA-Briefcase v1.1Artificial Analysis457–
BrowseCompOpenAI31.2%–
DeepSearchQAGoogle60.2%63.7%†
FACTS SearchGoogle90.2%–
Maths
FrontierMath Tiers 1–3Epoch AI Benchmarking Hub59.6%78.6%
FrontierMath Tier 4Epoch AI Benchmarking Hub26.8%49.0%
OTIS Mock AIMEEpoch AI Benchmarking Hub95.6%95.3%
ProofBenchVals AI26.0%56.0%
LiveBench MathematicsLiveBench91.0%†94.2%
AIME (Vals)Vals AI98.1%†96.7%
AIME 2026MathArena98.3%99.2%
HMMT February 2026MathArena94.7%97.7%
USAMO 2026MathArena74.4%95.2%
MathArena ApexMathArena60.9%54.2%
Arena MathsArena (LMArena)14881460†
Knowledge
SimpleQA VerifiedEpoch AI Benchmarking Hub73.5%†45.1%
LiveBench 1.0 Data AnalysisLiveBench78.5%†79.3%
MMLU-ProVals AI91.0%†87.5%
LegalBenchVals AI87.4%†86.0%
CorpFinVals AI64.5%†65.3%
TaxEvalVals AI72.9%†74.0%
MedQAVals AI96.4%†96.1%
PRBench FinanceScale AI SEAL41.9%45.6%†
PRBench LegalScale AI SEAL44.0%44.4%†
MultiNRCScale AI SEAL64.7%58.3%†
BioMysteryBenchVals AI60.0%†–
CaseLaw v2Vals AI64.8%†63.8%
Excel Modeling BenchmarkVals AI52.6%†–
MedCodeVals AI59.1%†41.3%
MedScribeVals AI76.1%†77.5%
Public Benefits Bench v1.1Vals AI53.8%†–
DrugDiscoveryBenchScale AI SEAL41.9%†–
EBR-benchEpoch AI Benchmarking Hub14.3%25.4%
GDP.pdfEpoch AI Benchmarking Hub17.0%–
Global-MMLU-Lite (AA)Artificial Analysis93.2%–
Instruction following
LiveBench LanguageLiveBench85.4%†82.6%
LiveBench Instruction FollowingLiveBench79.1%†70.2%
IFBenchArtificial Analysis77.1%74.0%
MultiChallengeScale AI SEAL71.4%–
TutorBenchScale AI SEAL53.0%–
Arena Instruction FollowingArena (LMArena)14801461†
Human preference
Arena TextArena (LMArena)14871465†
EQ-Bench 4EQ-Bench11321258†
Arena Creative WritingArena (LMArena)14811434†
Arena Multi-turnArena (LMArena)14961480†
Arena Longer QueriesArena (LMArena)14991477†
Multimodal
Arena VisionArena (LMArena)12961292†
MMMU-ProArtificial Analysis82.4%78.4%
VISTAScale AI SEAL–50.9%†
MMMU-Pro (official)MMMU80.5%†–
MMMU-Pro (Vals)Vals AI88.2%†87.5%
MortgageTaxVals AI69.4%†68.3%
SAGEVals AI83.2%†–
VTBScale AI SEAL29.0%29.2%†
Blueprint-Bench 2Epoch AI Benchmarking Hub26.5%27.1%†
Long context
AA-LCR v1.1Artificial Analysis82.0%82.0%
Arena DocumentArena (LMArena)14591470†
MLCRArtificial Analysis11.1%–
Composite
Epoch Capabilities IndexEpoch AI Benchmarking Hub156.2157.8†
LiveBench 1.0LiveBench77.0%†78.0%
AA Intelligence Index v4.3.2Artificial Analysis29.739.0
AA Progress Index v1Artificial Analysis296.1311.6
Vals RSI IndexVals AI–15.5%
Vals Science IndexVals AI27.4%†–
Vals Multimodal IndexVals AI56.1%†–
Safety & honesty
FORTRESSScale AI SEAL29.8%–
MASKScale AI SEAL42.4%89.7%†

Data as of 2026-11-06. Best configuration of each model; every score links to its source on the model pages.

DuelLab GameBench 2, release gb2-20261103-09b2218cbd85 (https://benchmarks.duellab.org/releases/gb2-20261103-09b2218cbd85/), CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). Scores mapped to BenchLeader model names; not endorsed by DuelLab.

What changed

  • GPT-5.4 (xhigh): BenchLeader Index 63.7 → 63.2.
  • Gemini 3.1 Pro: 1 score revision.
  • Gemini 3.1 Pro: 2 score revisions.
  • GPT-5.4 (xhigh): 1 score revision.
  • Gemini 3.1 Pro: BenchLeader Index 61.6 → 60.5.
  • Gemini 3.1 Pro: 2 new results (AA Progress Index v1, HLE-Diamond (AA)).
  • GPT-5.4 (xhigh): 1 new result (AA Progress Index v1).
  • Gemini 3.1 Pro: BenchLeader Index 59.3 → 61.6.
  • Gemini 3.1 Pro: Gemini 3.1 Pro cost per task moved from $1.30/M to $1.44/M (11% up).
  • Gemini 3.1 Pro: 2 score revisions.

Rankings: Best LLM overall · Best LLM for agents · Best cheap LLMs

Gemini 3.1 Pro vs GPT-5.4: questions

Is Gemini 3.1 Pro better than GPT-5.4?
Gemini 3.1 Pro and GPT-5.4 (xhigh) are neck and neck on quality (60.5 vs 63.2, within the ±4.9 margin). The BenchLeader Index combines every independent quality benchmark; the two are within the margin of error as of 6 Nov 2026, so check the category scores for your use.
Gemini 3.1 Pro vs GPT-5.4: which is cheaper, and will that change?
Gemini 3.1 Pro is cheaper today: $4.50 against $5.63 per million tokens, blended at three input tokens per output token. No price change has been announced for either.
Is Gemini 3.1 Pro or GPT-5.4 better for coding?
GPT-5.4 scores higher in coding (65 vs 57 on the category index, where 50 is average). On SciCode, the top shared coding benchmark, it is 58.7% for Gemini 3.1 Pro against 56.6% for GPT-5.4.
Which is faster, Gemini 3.1 Pro or GPT-5.4?
Gemini 3.1 Pro streams faster (122 vs 64 output tokens per second, Artificial Analysis 7-day median).
Is Gemini 3.1 Pro or GPT-5.4 better for agents and tool use?
GPT-5.4 scores higher in agentic tasks (63 vs 57 on the category index, where 50 is average). On τ³-Banking (AA), the top shared agents & tools benchmark, it is 21.4% for Gemini 3.1 Pro against 39.6% for GPT-5.4.
More questions (4)
Which is better for long documents?
They are level on long-context benchmarks (66 vs 66). Both take about 1.1M tokens: Gemini 3.1 Pro 1,048,576, GPT-5.4 1,050,000.
Which has lower latency?
Gemini 3.1 Pro answers first: 18.2 s against 147.3 s to the first answer token, as Artificial Analysis measures it. Reasoning effort drives most of this: a higher setting thinks longer before it answers.
Do independent benchmarks agree on Gemini 3.1 Pro vs GPT-5.4?
Across the 35 benchmarks both have, Gemini 3.1 Pro leads 13, GPT-5.4 leads 21 and 1 is a tie. Every overall index that rates both puts Gemini 3.1 Pro ahead.
When was GPT-5.4 released?
GPT-5.4 was released on 5 Mar 2026.