Inference
EN
OpenAI

GPT-5.4 nano

GPT-5.4 nano is a reasoning model from OpenAI in the GPT-5.4 family. 43 benchmarks count toward its index, in 8 categories.

  • Closed weights
  • Reasoning
  • Input: file · image · text
  • Released 2026-03
RankedEnough results for an index.
IndexAll categories, out of 100.50.9±2.5Median 51.0
CoverageIndex weight that has results.100%
SpeedOutput tokens per second.91/sMedian 67/s
First tokenSeconds to the first output token.0.9 sMedian 1.3 s
Input / 1MUS dollars per 1M input tokens.$0.20batch $0.10Median $0.60
Output / 1MUS dollars per 1M output tokens.$1.25batch $0.625 Batch price per 1M output tokens.Median $2.55
ContextMaximum tokens in one request.400KMedian 262K
EloLMArena rating and rank.1372±4 #187Median 1422

Details

From OpenRouter
API IDThe model ID on OpenRouter.
openai/gpt-5.4-nano
Max outputMaximum output tokens per request.
128K
Knowledge cutoffLast date of training data.
31 Aug 2025
ToolsTool calls on OpenRouter.
Yes
JSON outputOutput follows a given JSON schema.
Yes
Cache read / 1MUS dollars per 1M cached tokens.
$0.02
WeightsLink to the published weights.
—
RetiresDate OpenRouter removes the model.
—

Where to run

APIs that serve this model.
4 providers
ProviderTypeMaker, cloud, host or router.InputOutputContext
DatabricksCloud$0.20$1.25400K
DigitalOceanCloud$0.20$1.25400K
Microsoft AzureCloud$0.20$1.25400K
OpenRouterRouter$0.20$1.25400K

US dollars per 1M tokens, list price. From models.dev (opens in a new tab) (MIT).

Capabilities

Index per category, out of 100.
Out of 100
AgenticMulti-step tasks with tools.10 counted50.5125 / 218
CodingCode writing and repair.7 counted53.7109 / 209
ReasoningLogic problems and puzzles.7 counted53.791 / 227
MultimodalTasks with images and text.3 counted41.774 / 84
KnowledgeFacts and expert knowledge.6 counted44.5170 / 268
MultilingualTasks in many languages.1 counted44.4—
InstructionTasks with strict prompt rules.2 counted61.312 / 63
MathMath problems.7 counted55.961 / 168

Results

43 counted
BenchmarkThe benchmark name.ResultThe score from the source.PlacePlace among results on this benchmark.
Artificial Analysis IFBenchInstruction75.9%16 / 146
EuroEval SwedishMultilingual59.1% ±0.636 / 195
EuroEval SpanishMultilingual53.4% ±0.433 / 176
EuroEval SpanishMultilingual53.3% ±0.534 / 176
EuroEval PolishMultilingual55.2% ±0.541 / 209
EuroEval SwedishMultilingual57.8% ±0.542 / 195
EuroEval GermanMultilingual49.2% ±0.642 / 191
τ²-Bench Tool-Agent-User EvaluationAgentic92.5%34 / 149
EuroEval PortugueseMultilingual56.7% ±0.542 / 182
EuroEval SpanishMultilingual50.8% ±0.541 / 176
EuroEval GermanMultilingual48.7% ±0.645 / 191
EuroEval SwedishMultilingual56.4% ±0.748 / 195
EuroEval PortugueseMultilingual55.8% ±0.546 / 182
EuroEval ItalianMultilingual55.7% ±0.545 / 174
EuroEval ItalianMultilingual55.1% ±0.548 / 174
EuroEval PolishMultilingual52.4% ±0.659 / 209
EuroEval GermanMultilingual46.5% ±0.654 / 191
LiveCodeBenchCoding84.0% ±1.041 / 142
AIMEMath88.8% ±1.028 / 95
EuroEval PolishMultilingual50.7% ±0.463 / 209
Chess PuzzlesReasoning45.0% ±5.028 / 90
EuroEval DutchMultilingual53.1% ±0.455 / 175
EuroEval ItalianMultilingual52.5% ±0.555 / 174
EuroEval DutchMultilingual53.0% ±0.556 / 175
Critical Physics TasksReasoning9.3%69 / 214
EuroEval PortugueseMultilingual52.7% ±0.560 / 182
OTIS Mock AIME 2024-2025Math87.8% ±3.968 / 200
FrontierMath-2025-02-28-PrivateMath25.9% ±2.624 / 69
EuroEval FrenchMultilingual54.9% ±0.564 / 178
EuroEval DutchMultilingual51.6% ±0.463 / 175
EuroEval FrenchMultilingual54.8% ±0.566 / 178
Artificial Analysis Long Context ReasoningReasoning76.7%84 / 217
Artificial Analysis Humanity's Last ExamKnowledge28.3%90 / 222
FrontierMath-Tier-4-2025-07-01-PrivateMath6.3% ±3.523 / 55
EuroEval FrenchMultilingual53.0% ±0.575 / 178
LiveBench MathematicsMath91.0%28 / 66
Graduate-Level Google-Proof Q&AKnowledge82.8%61 / 141
GPQA diamond · Epoch AIKnowledge78.5% ±2.494 / 217
Artificial Analysis GPQA DiamondKnowledge81.7%93 / 199
Humanity's Last ExamKnowledge37.7%36 / 77
IOI v1Coding15.3% ±6.429 / 61
Terminal-Bench 2.0 · Vals AIAgentic39.9% ±4.534 / 66
LiveBench Instruction FollowingInstruction67.2%35 / 66
Artificial Analysis SciCodeCoding47.2%64 / 114
GPQA Diamond · Vals AIKnowledge77.5% ±2.476 / 135
APEX-Agents-AAAgentic24.9%17 / 30
ARC-AGI-1 (semi-private)Reasoning51.5%74 / 126
GDPval-AA normalizedAgentic22.4%79 / 133
ARC-AGI-2 (semi-private)Reasoning5.7%78 / 131
Artificial Analysis Omniscience AccuracyKnowledge25.7%102 / 164
MMMU ProMultimodal73.6% ±1.159 / 92
FrontierMath-Tiers-1-3-v2-PrivateMath44.9% ±3.059 / 90
Vibe Code Bench v1.1Coding26.1% ±5.183 / 123
LiveBench ReasoningReasoning81.1%55 / 80
Artificial Analysis MMMU-ProMultimodal65.4%82 / 113
Code MigrationCoding14.5% ±4.056 / 77
SWE-benchCoding69.8% ±2.164 / 87
LiveBench Agentic CodingAgentic46.8%49 / 66
LiveBench Data AnalysisReasoning67.6%61 / 80
MCP AtlasAgentic56.1%40 / 52
MMLU ProKnowledge77.2% ±0.4105 / 135
Terminal-Bench 2.0 · BenchLMAgentic46.3%33 / 41
FrontierMath-Tier-4-v2-PrivateMath12.2% ±5.258 / 72
ExtractBench: schema-guided enterprise document extraction (LlamaIndex leaderboard)Multimodal74.9%22 / 26
Massive Multi-discipline Multimodal Understanding ProMultimodal66.1%46 / 53
LiveBench CodingCoding70.8%58 / 66
Humanity's Last Exam without toolsKnowledge24.3%47 / 53
Terminal-Bench 2.1 · Vals AIAgentic41.6% ±3.068 / 75
ToolathlonAgentic35.5%20 / 22
ParseBench: document parsing for AI agents (LlamaIndex leaderboard)Multimodal42.2%36 / 39
Mystery Game PuzzlesReasoning5.0% ±2.275 / 80
SimpleQA VerifiedKnowledge11.7% ±1.087 / 91
LiveBench LanguageKnowledge62.5%64 / 66
OSWorld-VerifiedAgentic39.0%37 / 37
FrontierMath-2025-02-28-PublicMath60.0% ±16.311 / 64
CorpFin v2Finance61.2% ±1.055 / 131
MedCodeHealthcare41.0% ±2.359 / 106
Finance Agent v1.1Finance47.8% ±2.830 / 51
ARC-AGI-2 (public eval)Reasoning4.2%74 / 122
MedScribeHealthcare77.1% ±1.971 / 109
ARC-AGI-1 (public eval)Reasoning47.9%79 / 116
MortgageTaxFinance59.1% ±1.068 / 97
LegalBenchLegal77.9% ±0.4105 / 149
TaxEval v2Finance67.4% ±0.9107 / 143
EMBFinance44.8% ±3.159 / 75
Finance Agent (v2)Finance38.2% ±1.264 / 79
CaseLaw v2Legal51.9% ±1.145 / 53
Legal Research BenchLegal6.3% ±1.784 / 90
Tax Agent BenchFinance5.0% ±1.478 / 82
MMMU-Pro with PythonMultimodal69.5%—

Sources

BenchLMCC BY-NC 4.0OpenRouterTerms restrict reuseLiveBenchNo licence statedVals AINo licence statedEpoch AICC BYEuroEvalMITARC Prize FoundationTerms restrict reusemodels.devMIT

History

5 releases
Index51.8 → 50.9
23 Oct 202612 Nov 2026
Output / 1M$1.25 → $1.25
23 Oct 202612 Nov 2026

Questions

How much does GPT-5.4 nano cost?
$0.20 per 1M input tokens, $1.25 per 1M output tokens.
How good is GPT-5.4 nano?
Its index is 50.9 out of 100. That is place 143 of 283 ranked models.
What is GPT-5.4 nano best at?
Instruction: 61.3 out of 100, place 12 of 63.
How fast is GPT-5.4 nano?
91 output tokens per second, 0.9 s to the first token.
What is the context window of GPT-5.4 nano?
400K tokens.
Is GPT-5.4 nano open source?
No. Its weights are not published.
When was GPT-5.4 nano released?
March 2026.

Badge

GPT-5.4 nano on Inference360: place 143 of 283, index 50.9
HTML
<a href="https://inference360.ai/models/gpt-5-4-nano"><img src="https://inference360.ai/badges/gpt-5-4-nano.svg" alt="GPT-5.4 nano on Inference360: place 143 of 283, index 50.9" height="20"></a>
Markdown
[![GPT-5.4 nano on Inference360: place 143 of 283, index 50.9](https://inference360.ai/badges/gpt-5-4-nano.svg)](https://inference360.ai/models/gpt-5-4-nano)

Compare with

Ling 3.0 Flash50.9Nemotron 3.5 Super VL 120B A12B51.0Nemotron 3.5 Super VL 120B A12B NVFP450.8MiniMax M2.751.2GPT-5.4 mini54.5

Lower price

dots3-note Preview63.4 · FreeApodex 1.1 Mini55.7 · FreeNemotron 3 Ultra50.8 · Free

More from OpenAI

All models
GPT-5.4 mini54.5GPT-5.4 Pro73.3GPT-5.464.8GPT-5.3 InstantUnrankedGPT Realtime 1.5Unranked