Skip to main content
BenchLM
Advertise

Model profile

GPT-5.4 mini

OpenAICurrentReleased Mar 17, 2026
Data verified
Overall Score
56.77Public #75 of 200
Arena Elo
1449
Eligible category ranks
3of 8
Price (1M tokens)
$0.75 in / $4.5 out
API pricing
Speed
201tok/s
Context
400K

Evidence coverage

30 of 323 tracked benchmarks are published. 13 are verified and 17 provisional. 7 of 8 categories are measured.

Updated July 23, 2026Methodology
Published / tracked
30 / 323
Verified
13
Provisional
17
Categories with evidence
7 / 8

Evidence by category

  • Agentic9 benchmarks
    Mixed evidence
  • Coding4 benchmarks
    Mixed evidence
  • Reasoning2 benchmarks
    Reported
  • Knowledge9 benchmarks
    Mixed evidence
  • Math2 benchmarks
    Verified
  • Multilingual0 benchmarks
    Not measured
  • Multimodal3 benchmarks
    Mixed evidence
  • Inst. Following1 benchmark
    Reported
ProprietaryReasoning
Confidence:
High
mini

GPT-5.4 mini ranks #75 out of 200 models on the public leaderboard with an overall score of 56.77/100. It does not yet have enough sourced coverage for BenchLM's verified leaderboard. While not a frontier model, it offers specific advantages depending on the use case.

GPT-5.4 mini is a proprietary model with a 400K token context window. It uses explicit chain-of-thought reasoning, which typically improves performance on math and complex reasoning tasks at the cost of higher latency and token usage.

GPT-5.4 mini sits inside the GPT-5.4 family alongside GPT-5.4, GPT-5.4 nano, GPT-5.4 Pro. BenchLM links it directly to GPT-5 mini as the earlier related model in that lineage. This profile currently has 30 of 323 tracked benchmarks. BenchLM only exposes non-generated benchmark rows publicly, so missing categories stay blank until a sourced evaluation is available.

Its strongest category is Knowledge (#37), while its weakest is Agentic (#100). This performance profile makes it particularly effective for knowledge-intensive tasks like research, analysis, and factual Q&A.

Peer position

Exact provisional scores and ranks for the closest listed peers. A score can appear before a model clears the evidence threshold for a rank, so equal scores can have different rank states.

Range 56.58–57.01

  1. Qwen3.5 397B
    Alibaba
    #7157.01
    Qwen3.5 397B is #71 with a score of 57.01.
    Compare
  2. Qwen3.5-35B-A3B
    Alibaba
    #7256.97
    Qwen3.5-35B-A3B is #72 with a score of 56.97.
    Compare
  3. GPT-5.3-Codex-Spark
    OpenAI
    #7356.91
    GPT-5.3-Codex-Spark is #73 with a score of 56.91.
    Compare
  4. Kimi K2.6
    Moonshot AI
    #7456.79
    Kimi K2.6 is #74 with a score of 56.79.
    Compare
  5. GPT-5.4 miniCurrent model
    OpenAI
    #7556.77
    GPT-5.4 mini is #75 with a score of 56.77.
  6. Grok 4 Fast (Reasoning)
    xAI
    #7656.59
    Grok 4 Fast (Reasoning) is #76 with a score of 56.59.
    Compare
  7. Claude Haiku 4.5
    Anthropic
    #7756.58
    Claude Haiku 4.5 is #77 with a score of 56.58.
    Compare

Category percentile

More

Relative position among models eligible for each sourced category. A higher percentile means a stronger position within that category's ranked cohort; 100 is highest.

  1. Knowledge29%
    Eligible cohort rank #37 of 52Category score 63.4
  2. Coding26%
    Eligible cohort rank #90 of 122Category score 46.0
  3. Agentic16%
    Eligible cohort rank #100 of 119Category score 37.5

Category evidence

Scores and ranks appear only where this model has published benchmark evidence. Categories without displayable source records remain not measured.

Category scores, ranks, weighting, benchmark coverage, and evidence status
CategoryScore
AgenticRank #100 of 119Percentile 16thWeight 22%9 benchmarksMixed sources37.5
CodingRank #90 of 122Percentile 26thWeight 20%4 benchmarksMixed sources46.0
ReasoningRank Not rankedWeight 17%2 benchmarksReported40.7
KnowledgeRank #37 of 52Percentile 29thWeight 12%9 benchmarksMixed sources63.4
MathRank Not rankedWeight 5%2 benchmarksVerified44.7
MultilingualWeight 7%0 benchmarksNot measuredNot measured
MultimodalRank Not rankedWeight 12%3 benchmarksMixed sources58.2
Inst. FollowingRank Not rankedWeight 5%1 benchmarkReported87.4

Chatbot Arena performance

Scroll horizontally to inspect confidence intervals and vote counts.

Chatbot Arena Elo, confidence interval, and vote count by evaluation view
ViewEloConfidence intervalVotes
Text Overall1449±3.956,941
Coding1496±6.215,508
Math1438±11.33,060
Instruction Following1435±5.718,980
Creative Writing1405±7.59,308
Multi-turn1471±7.010,611
Hard Prompts1471±4.737,035
Hard Prompts (English)1468±5.918,266
Longer Query1449±5.623,961

Benchmark Details

Rows below have a displayable published verification record. Each source link and provenance note remains in the page HTML while its category is closed. Source-unverified manual rows and generated rows stay hidden.

Agentic9 benchmarks
Terminal-Bench 2.0Provider exact
60%Weighted 38%
Source: OpenAI: Introducing GPT-5.4 mini and nanoProvenance: Provider exact
OSWorld-VerifiedProvider exact
72.1%Weighted 34%
Source: OpenAI: Introducing GPT-5.4 mini and nanoProvenance: Provider exact
MCP AtlasProvider exact
57.7%Display only
Source: OpenAI: Introducing GPT-5.4 mini and nanoProvenance: Provider exact
ToolathlonProvider exact
42.9%Display only
Source: OpenAI: Introducing GPT-5.4 mini and nanoProvenance: Provider exact
τ²-bench resultsProvider exact

τ²-Bench Tool-Agent-User Evaluation

83.3%Display only
Source: OpenAI: Introducing GPT-5.4 mini and nanoProvenance: OpenAI reports GPT-5.4 mini at 93.4% on τ²-bench Telecom at xhigh reasoning effort.
AA Agentic IndexReported

Artificial Analysis Agentic Index

30.2%Display only
Source: Artificial Analysis model benchmarksProvenance: Display-only row synced from the current Artificial Analysis model payload. It is excluded from BenchLM weighted scoring.
APEX-Agents-AAReported
28.2%Display only
Source: Artificial Analysis model benchmarksProvenance: Display-only row synced from the current Artificial Analysis model payload. It is excluded from BenchLM weighted scoring.
GDPval-AAReported

GDPval-AA normalized

33.6%Display only
Source: Artificial Analysis model benchmarksProvenance: Display-only row synced from the current Artificial Analysis model payload. It is excluded from BenchLM weighted scoring.
GDPval-AAReported
1171Display only
Source: Artificial Analysis model benchmarksProvenance: Display-only row synced from the current Artificial Analysis model payload. It is excluded from BenchLM weighted scoring.
Coding4 benchmarks
Vibe Code BenchBenchmark exact

Vibe Code Bench v1.1

47.97%Display only
Source: Vals AI: Vibe Code Bench v1.1Provenance: Vals Vibe Code Bench v1.1 reports this exact row under openai/gpt-5.4-mini-2026-03-17; BenchLM stores it on the local vibeCodeBench key.
AA Coding IndexReported

Artificial Analysis Coding Index

56.1%Display only
Source: Artificial Analysis model benchmarksProvenance: Display-only row synced from the current Artificial Analysis model payload. It is excluded from BenchLM weighted scoring.
AA-SciCodeReported

Artificial Analysis SciCode

49.9%Display only
Source: Artificial Analysis model benchmarksProvenance: Display-only row synced from the current Artificial Analysis model payload. It is excluded from BenchLM weighted scoring.
FrontierCode 1.1 MainBenchmark exact
27.0%Display only
Source: Cognition: FrontierCode 1.1Provenance: Cognition reports GPT-5.4 mini at 27.0% on FrontierCode 1.1 Main, using the best xhigh effort row from the published data JSON.
Reasoning2 benchmarks
AA-LCRReported

Artificial Analysis Long Context Reasoning

69.3%Display only
Source: Artificial Analysis model benchmarksProvenance: Display-only row synced from the current Artificial Analysis model payload. It is excluded from BenchLM weighted scoring.
CritPtReported

Critical Physics Tasks

10.0%Display only
Source: Artificial Analysis model benchmarksProvenance: Display-only row synced from the current Artificial Analysis model payload. It is excluded from BenchLM weighted scoring.
Knowledge9 benchmarks
HLEProvider exact

Humanity's Last Exam

41.5%Weighted 45%
Source: OpenAI: Introducing GPT-5.4 mini and nanoProvenance: Provider exact
GPQAReported

Graduate-Level Google-Proof Q&A

88%Weighted 7%
Source: Reported upstream sourceProvenance: Reported row carried from an upstream public source. Displayable on BenchLM, but not treated as verified unless explicitly marked otherwise.
HLE w/o toolsProvider exact

Humanity's Last Exam without tools

28.2%Display only
Source: OpenAI: Introducing GPT-5.4 mini and nanoProvenance: Provider exact
Artificial Analysis Intelligence IndexReported
40.0%Display only
Source: Artificial Analysis model benchmarksProvenance: Display-only row synced from the current Artificial Analysis model payload. It is excluded from BenchLM weighted scoring.
AA-GPQA DiamondReported

Artificial Analysis GPQA Diamond

87.5%Display only
Source: Artificial Analysis model benchmarksProvenance: Display-only row synced from the current Artificial Analysis model payload. It is excluded from BenchLM weighted scoring.
AA-HLEReported

Artificial Analysis Humanity's Last Exam

26.6%Display only
Source: Artificial Analysis model benchmarksProvenance: Display-only row synced from the current Artificial Analysis model payload. It is excluded from BenchLM weighted scoring.
AA-Omniscience IndexReported

Artificial Analysis Omniscience Index

-18.7%Display only
Source: Artificial Analysis model benchmarksProvenance: Display-only row synced from the current Artificial Analysis model payload. It is excluded from BenchLM weighted scoring.
AA-Omniscience AccuracyReported

Artificial Analysis Omniscience Accuracy

37.5%Display only
Source: Artificial Analysis model benchmarksProvenance: Display-only row synced from the current Artificial Analysis model payload. It is excluded from BenchLM weighted scoring.
AA-Omniscience Hallucination RateReported

Artificial Analysis Omniscience Hallucination Rate

89.8%Display only
Source: Artificial Analysis model benchmarksProvenance: Display-only row synced from the current Artificial Analysis model payload. It is excluded from BenchLM weighted scoring.
Math2 benchmarks
FrontierMath v2 (Tiers 1-3)Benchmark exact

FrontierMath v2 Tiers 1-3

28.280%Weighted 30%
Source: Epoch AI FrontierMath v2 leaderboardProvenance: Epoch AI reports FrontierMath v2 Tiers 1-3 at 28.28% for gpt-5.4-mini-2026-03-17_high. BenchLM selects the highest published thinking effort for the model and stores the v2 benchmark slice separately.
FrontierMath v2 (Tier 4)Benchmark exact

FrontierMath v2 Tier 4

2.080%Weighted 10%
Source: Epoch AI FrontierMath v2 leaderboardProvenance: Epoch AI reports FrontierMath v2 Tier 4 at 2.08% for gpt-5.4-mini-2026-03-17_high. BenchLM selects the highest published thinking effort for the model and stores the v2 benchmark slice separately.
Multimodal3 benchmarks
MMMU-ProProvider exact

Massive Multi-discipline Multimodal Understanding Pro

76.6%Weighted 45%
Source: OpenAI: Introducing GPT-5.4 mini and nanoProvenance: Provider exact
MMMU-Pro w/ PythonProvider exact

MMMU-Pro with Python

78%Display only
Source: OpenAI: Introducing GPT-5.4 mini and nanoProvenance: Provider exact
AA-MMMU-ProReported

Artificial Analysis MMMU-Pro

73.3%Display only
Source: Artificial Analysis model benchmarksProvenance: Display-only row synced from the current Artificial Analysis model payload. It is excluded from BenchLM weighted scoring.
Inst. Following1 benchmark
AA-IFBenchReported

Artificial Analysis IFBench

73.3%Display only
Source: Artificial Analysis model benchmarksProvenance: Display-only row synced from the current Artificial Analysis model payload. It is excluded from BenchLM weighted scoring.

GPT-5.4 Family

Mini

Canonical Entry

GPT-5.4

Related Earlier Model

GPT-5 mini
GPT-5.4Score 74.24GPT-5.4 nanoScore 66.79GPT-5.4 ProScore 60.89

Frequently Asked Questions

How does GPT-5.4 mini perform overall in AI benchmarks?

GPT-5.4 mini currently ranks #75 out of 200 models on BenchLM's provisional leaderboard with an overall score of 56.77. It is created by OpenAI. Its published context window is 400K.

Is GPT-5.4 mini good for knowledge and understanding?

GPT-5.4 mini ranks #37 out of 52 models in knowledge and understanding benchmarks with an average score of 63.4. There are stronger options in this category.

Is GPT-5.4 mini good for coding and programming?

GPT-5.4 mini ranks #90 out of 122 models in coding and programming benchmarks with an average score of 46. There are stronger options in this category.

Is GPT-5.4 mini good for mathematics?

GPT-5.4 mini has visible benchmark coverage in mathematics, but BenchLM does not currently assign it a global category rank there.

Is GPT-5.4 mini good for reasoning and logic?

GPT-5.4 mini has visible benchmark coverage in reasoning and logic, but BenchLM does not currently assign it a global category rank there.

Is GPT-5.4 mini good for agentic tool use and computer tasks?

GPT-5.4 mini ranks #100 out of 119 models in agentic tool use and computer tasks benchmarks with an average score of 37.5. There are stronger options in this category.

Is GPT-5.4 mini good for multimodal and grounded tasks?

GPT-5.4 mini has visible benchmark coverage in multimodal and grounded tasks, but BenchLM does not currently assign it a global category rank there.

Is GPT-5.4 mini good for instruction following?

GPT-5.4 mini has visible benchmark coverage in instruction following, but BenchLM does not currently assign it a global category rank there.

Which sibling models are related to GPT-5.4 mini?

GPT-5.4 mini belongs to the GPT-5.4 family. Related variants on BenchLM include GPT-5.4, GPT-5.4 nano, GPT-5.4 Pro.

Does GPT-5.4 mini have full benchmark coverage on BenchLM?

Not yet. GPT-5.4 mini currently has 30 published benchmark scores out of the 323 benchmarks BenchLM tracks. BenchLM only exposes non-generated public benchmark rows, so missing categories stay blank until a sourced evaluation is available.

What is the context window size of GPT-5.4 mini?

GPT-5.4 mini has a published context window of 400K, which determines how much text it can process in a single interaction.

Related Resources

More from OpenAI LLM Pricing Comparison BenchLM Methodology The AI Race Which LLM Should I Use? Benchmark Explainers
Last updated: July 23, 2026 · Runtime metrics stay blank until BenchLM has a sourced snapshot.

Choose with this week’s evidence

Join 2,000+ readers for ranking moves, new releases, pricing changes, and the evidence behind them.

Free. One email per week.