Skip to main content

Model profile

GPT-5.2

OpenAIEstablishedReleased Dec 11, 2025
Data verified
Overall Score
58.43Public #64 of 200
Arena Elo
1435
Eligible category ranks
4of 8
Price (1M tokens)
$1.75 in / $14 out
API pricing
Speed
73tok/s
Context
400K

Evidence coverage

28 of 321 tracked benchmarks are published. 11 are verified and 17 provisional. 7 of 8 categories are measured.

Updated July 20, 2026Methodology
Published / tracked
28 / 321
Verified
11
Provisional
17
Categories with evidence
7 / 8

Evidence by category

  • Agentic5 benchmarks
    Mixed evidence
  • Coding4 benchmarks
    Mixed evidence
  • Reasoning3 benchmarks
    Mixed evidence
  • Knowledge7 benchmarks
    Reported
  • Math3 benchmarks
    Mixed evidence
  • Multilingual0 benchmarks
    Not measured
  • Multimodal5 benchmarks
    Mixed evidence
  • Inst. Following1 benchmark
    Reported
ProprietaryReasoning
Confidence:
High
thinking

GPT-5.2 ranks #64 out of 200 models on the public leaderboard with an overall score of 58.43/100. 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-5.2 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-5.2 sits inside the GPT-5.2 family alongside GPT-5.2 Instant, GPT-5.2 Pro. This profile currently has 28 of 321 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 Multimodal & Grounded (#12), while its weakest is Agentic (#90). This performance profile makes it particularly strong for screenshots, documents, charts, and grounded multimodal workflows.

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 58.0158.9

  1. GPT-5.3 Instant
    OpenAI
    #6058.9
    GPT-5.3 Instant is #60 with a score of 58.9.
    Compare
  2. DeepSeek V4 Flash
    DeepSeek
    #6158.88
    DeepSeek V4 Flash is #61 with a score of 58.88.
    Compare
  3. MiMo-V2.5
    Xiaomi
    #6258.62
    MiMo-V2.5 is #62 with a score of 58.62.
    Compare
  4. GPT-5 (high)
    OpenAI
    #6358.61
    GPT-5 (high) is #63 with a score of 58.61.
    Compare
  5. GPT-5.2Current model
    OpenAI
    #6458.43
    GPT-5.2 is #64 with a score of 58.43.
  6. DeepSeek V3.2 (Thinking)
    DeepSeek
    #6558.15
    DeepSeek V3.2 (Thinking) is #65 with a score of 58.15.
    Compare
  7. Qwen3 235B 2507 (Reasoning)
    Alibaba
    #6658.01
    Qwen3 235B 2507 (Reasoning) is #66 with a score of 58.01.
    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. Knowledge78%
    Eligible cohort rank #12 of 52Category score 81.0
  2. Multimodal61%
    Eligible cohort rank #12 of 29Category score 69.3
  3. Coding70%
    Eligible cohort rank #37 of 122Category score 54.2
  4. Agentic25%
    Eligible cohort rank #90 of 119Category score 41.6

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 #90 of 119Percentile 25thWeight 22%5 benchmarksMixed sources41.6
CodingRank #37 of 122Percentile 70thWeight 20%4 benchmarksMixed sources54.2
ReasoningRank Not rankedWeight 17%3 benchmarksMixed sources65.4
KnowledgeRank #12 of 52Percentile 78thWeight 12%7 benchmarksReported81.0
MathRank Not rankedWeight 5%3 benchmarksMixed sources58.5
MultilingualWeight 7%0 benchmarksNot measuredNot measured
MultimodalRank #12 of 29Percentile 61stWeight 12%5 benchmarksMixed sources69.3
Inst. FollowingRank Not rankedWeight 5%1 benchmarkReported82.4

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 Overall1435±3.375,293
Coding1482±5.419,600
Math1433±9.74,166
Instruction Following1422±5.024,212
Creative Writing1390±6.512,208
Multi-turn1444±6.014,241
Hard Prompts1459±4.147,089
Hard Prompts (English)1458±5.123,340
Longer Query1441±5.029,033

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.

Agentic5 benchmarks
OSWorld-VerifiedProvider exact
47.3%Weighted 34%
Source: OpenAI: Introducing GPT-5.4Provenance: Provider exact
BrowseCompProvider exact
65.8%Weighted 28%
Source: OpenAI: Introducing GPT-5.4Provenance: Provider exact
τ²-bench resultsReported

τ²-Bench Tool-Agent-User Evaluation

84.8%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.
Gert LabsBenchmark exact

Gert Labs Composite Game Benchmark

46.54%Display only
Source: Gert Labs rankingsProvenance: Gert Labs reports this composite leaderboard score in the public rankings API. BenchLM scales the source gscore from 0-1 to 0-100 and stores it as a display-only agentic benchmark.
JobBenchBenchmark exact
34.3%Display only
Source: JobBench paperProvenance: JobBench reports GPT-5.2 under Codex CLI on the main-set leaderboard. BenchLM stores the reported main-set score.
Coding4 benchmarks
SWE-bench VerifiedProvider exact

Software Engineering Benchmark Verified

80%Weighted 16%
Source: OpenAI: Introducing GPT-5.2Provenance: OpenAI reports GPT-5.2 Thinking at 80.0% on SWE-bench Verified in the detailed coding benchmarks table.
SWE-bench ProProvider exact
55.6%Weighted 10%
Source: OpenAI: Introducing GPT-5.4Provenance: Provider exact
Vibe Code BenchBenchmark exact

Vibe Code Bench v1.1

53.50%Display only
Source: Vals AI: Vibe Code Bench v1.1Provenance: Vals Vibe Code Bench v1.1 reports this exact row under openai/gpt-5.2-2025-12-11; BenchLM stores it on the local vibeCodeBench key.
AA-SciCodeReported

Artificial Analysis SciCode

52.1%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.
Reasoning3 benchmarks
ARC-AGI-2Provider exact

Abstraction and Reasoning Corpus for AGI v2

52.9%Weighted 31%
Source: OpenAI: Introducing GPT-5.2Provenance: OpenAI reports GPT-5.2 Thinking at 52.9% on ARC-AGI-2 (Verified) in the abstract reasoning benchmarks table.
AA-LCRReported

Artificial Analysis Long Context Reasoning

72.7%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.
CritPtReported

Critical Physics Tasks

11.6%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.
Knowledge7 benchmarks
GPQASecondary exact

Graduate-Level Google-Proof Q&A

92.4%Weighted 7%
Source: Qwen3.5 language comparison tableProvenance: Qwen3.5 public comparison table reports GPT5.2 at 92.4 on GPQA, matching the stored BenchLM row.
Artificial Analysis Intelligence IndexReported
42.2%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.
AA-GPQA DiamondReported

Artificial Analysis GPQA Diamond

90.3%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.
AA-HLEReported

Artificial Analysis Humanity's Last Exam

35.4%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

-1.0%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-Omniscience AccuracyReported

Artificial Analysis Omniscience Accuracy

43.8%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.
AA-Omniscience Hallucination RateReported

Artificial Analysis Omniscience Hallucination Rate

79.7%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.
Math3 benchmarks
FrontierMath v2 (Tiers 1-3)Benchmark exact

FrontierMath v2 Tiers 1-3

40.700%Weighted 30%
Source: Epoch AI FrontierMath v2 leaderboardProvenance: Epoch AI reports FrontierMath v2 Tiers 1-3 at 40.7% for gpt-5.2-2025-12-11_xhigh. 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

18.800%Weighted 10%
Source: Epoch AI FrontierMath v2 leaderboardProvenance: Epoch AI reports FrontierMath v2 Tier 4 at 18.8% for gpt-5.2-2025-12-11_xhigh. BenchLM selects the highest published thinking effort for the model and stores the v2 benchmark slice separately.
AA AIME 2025Reported

Artificial Analysis AIME 2025

99.0%Display only
Source: Artificial Analysis: aime-2025 leaderboardProvenance: Display-only row synced from the current Artificial Analysis evaluation leaderboard. It is excluded from BenchLM weighted scoring.
Multimodal5 benchmarks
MMMU-ProProvider exact

Massive Multi-discipline Multimodal Understanding Pro

79.5%Weighted 45%
Source: OpenAI: Introducing GPT-5.4Provenance: Provider exact
CharXivSecondary exact

CharXiv Reasoning

82.1%Weighted 25%
Source: Qwen3.6-Plus multimodal comparison tableProvenance: Secondary exact
MathVisionSecondary exact
83.0%Display only
Source: Qwen3.6-Plus multimodal comparison tableProvenance: Secondary exact
V*Secondary exact
75.9%Display only
Source: Qwen3.6-Plus multimodal comparison tableProvenance: Qwen reports V* in the multimodal comparison table. BenchLM stores the published first value when a slash pair is provided.
Design Arena WebsiteReported

Design Arena Website Elo

1227Display only
Source: OpenRouter model benchmarksProvenance: Display-only Design Arena Website Elo synced from OpenRouter model benchmark metadata. It is excluded from BenchLM weighted scoring.
Inst. Following1 benchmark
AA-IFBenchReported

Artificial Analysis IFBench

75.4%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.

Frequently Asked Questions

How does GPT-5.2 perform overall in AI benchmarks?

GPT-5.2 currently ranks #64 out of 200 models on BenchLM's provisional leaderboard with an overall score of 58.43. It is created by OpenAI. Its published context window is 400K.

Is GPT-5.2 good for knowledge and understanding?

GPT-5.2 ranks #12 out of 52 models in knowledge and understanding benchmarks with an average score of 81. There are stronger options in this category.

Is GPT-5.2 good for coding and programming?

GPT-5.2 ranks #37 out of 122 models in coding and programming benchmarks with an average score of 54.2. There are stronger options in this category.

Is GPT-5.2 good for mathematics?

GPT-5.2 has visible benchmark coverage in mathematics, but BenchLM does not currently assign it a global category rank there.

Is GPT-5.2 good for reasoning and logic?

GPT-5.2 has visible benchmark coverage in reasoning and logic, but BenchLM does not currently assign it a global category rank there.

Is GPT-5.2 good for agentic tool use and computer tasks?

GPT-5.2 ranks #90 out of 119 models in agentic tool use and computer tasks benchmarks with an average score of 41.6. There are stronger options in this category.

Is GPT-5.2 good for multimodal and grounded tasks?

GPT-5.2 ranks #12 out of 29 models in multimodal and grounded tasks benchmarks with an average score of 69.3. There are stronger options in this category.

Is GPT-5.2 good for instruction following?

GPT-5.2 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-5.2?

GPT-5.2 belongs to the GPT-5.2 family. Related variants on BenchLM include GPT-5.2 Instant, GPT-5.2 Pro.

Does GPT-5.2 have full benchmark coverage on BenchLM?

Not yet. GPT-5.2 currently has 28 published benchmark scores out of the 321 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-5.2?

GPT-5.2 has a published context window of 400K, which determines how much text it can process in a single interaction.

Last updated: July 20, 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.