Noometry

Model comparison

DeepSeek-R1 vs GPT-5.2

GPT-5.2 is the stronger model overall, scoring 54.1 to 42.3 on the Noometry Index. DeepSeek-R1 costs 5.3× less per token, which makes it the better buy when GPT-5.2's lead doesn't matter for your workload.

Last verified . 32 shared benchmarks.

DeepSeek-R1 DeepSeek

42.3

Rank #115 Confirmed

GPT-5.2 OpenAI

54.1

Rank #34 Confirmed

Summary

  • They share 32 benchmarks with published results for both. DeepSeek-R1 scores higher in 1 category and GPT-5.2 in 8 categories; 8 gaps are clear of the uncertainty.
  • The widest gap is in reasoning, where GPT-5.2 leads 50.2 to 18.6.
  • The biggest single-benchmark swing is ARC-AGI-1: 21.2% for DeepSeek-R1 and 86.2% for GPT-5.2.
  • DeepSeek-R1 is cheaper at $0.50 / $2.15 per million input/output tokens, against $1.75 / $14 for GPT-5.2.
  • GPT-5.2 accepts more context: 400K tokens versus 164K.

Side by side

DeepSeek-R1GPT-5.2
ProviderDeepSeekOpenAI
Noometry Index42.354.1
Released2025-01-202025-12-11
WeightsProprietaryProprietary
Context window164K400K
Max output64K128K
Input $ / M tokens$0.50$1.75
Output $ / M tokens$2.15$14
Results tracked5267

Sponsored placements are available on pages like this one. Advertise on Noometry

Category by category

Coding GPT-5.2 leads

DeepSeek-R1: 46.3 (#68), GPT-5.2: 51.6 (#37)

BenchmarkDeepSeek-R1GPT-5.2
WeirdML41.6%72.2%
LMArena Coding14271447
ALE-Bench804.121,294
AlgoTune1.72.05
SWE-bench Verified—73.8%
SWE-bench Verified (bash only)—72.8%
Aider Polyglot71.4%—
LMArena WebDev—1416
SWE-bench Multilingual—66.7%
SciCode35.7%—
GSO—27.4%
LiveBench Coding66.7%—

Agentic & Tool Use GPT-5.2 leads

DeepSeek-R1: 30.7 (#75), GPT-5.2: 40.2 (#24)

BenchmarkDeepSeek-R1GPT-5.2
DeepResearch Bench35.1%41.1%
METR Time Horizons53.8%75.3%
Terminal-Bench—64.9%
Berkeley Function Calling Leaderboard—55.9%
GDPval—49.7%
Remote Labor Index—2.5%
τ²-bench Airline—83%
τ²-bench Banking—32.2%
τ²-bench Retail—81.6%
τ²-bench Telecom—89.7%
BALROG34.9%—
LMArena Search—1207
Vending-Bench 2—3,591

Reasoning GPT-5.2 leads

DeepSeek-R1: 18.6 (#278), GPT-5.2: 50.2 (#35)

BenchmarkDeepSeek-R1GPT-5.2
ARC-AGI-21.3%52.9%
SimpleBench40.8%45.8%
Kagi LLM Benchmark69.4%73.3%
ARC-AGI-121.2%86.2%
LMArena Hard Prompts14161445
Epoch Capabilities Index141.29153.45
ForecastBench6060.1
NYT Connections (extended)—83.6%
CritPt1.1%—
Chess Puzzles—49%
EnigmaEval—10.4%
EBR-Bench—23%
LiveBench Reasoning83.2%—
Mystery Game Puzzles—23%
DTBench—90.9%
LiveBench Data Analysis69.8%—
LMCA—43.9%
LiveBench71.6%—

Math GPT-5.2 leads

DeepSeek-R1: 43.8 (#79), GPT-5.2: 60.0 (#38)

BenchmarkDeepSeek-R1GPT-5.2
OTIS Mock AIME 2024-202566.4%96.1%
LMArena Math14001440
FrontierMath (Tiers 1-3)—67.4%
FrontierMath Tier 4—31.7%
MathArena Final-Answer Competitions—72%
ProofBench—15%
Omni-MATH42.4%—
LiveBench Math80.7%—
MATH Level 596.6%—
FrontierMath (Feb 2025 set)—40.7%
FrontierMath Tier 4 (v1)—18.8%

Knowledge GPT-5.2 leads

DeepSeek-R1: 44.5 (#87), GPT-5.2: 59.3 (#32)

BenchmarkDeepSeek-R1GPT-5.2
GPQA Diamond76.3%91.4%
Vectara Hallucination Rate11.3%8.4%
LMArena Expert13941445
Humanity's Last Exam—27.8%
SimpleQA Verified—37.1%
MMLU-Pro79.3%—
Confabulations12.7%—
GPQA (HELM)66.6%—

Multimodal Not comparable

DeepSeek-R1: —, GPT-5.2: 51.3 (#7)

BenchmarkDeepSeek-R1GPT-5.2
LMArena Vision—1268
VPCT—84%
Furniture Assembly—38.3%
LMArena Document—1405

Multilingual Too close to call

DeepSeek-R1: 52.4 (#85), GPT-5.2: 53.4 (#67)

BenchmarkDeepSeek-R1GPT-5.2
LMArena Non-English14121425
LMArena Chinese14421460
LMArena French14171455
LMArena German14041448
LMArena Japanese13911420
LMArena Korean13601392
LMArena Russian14231440
LMArena Spanish14111433

Instruction Following GPT-5.2 leads

DeepSeek-R1: 72.0 (#143), GPT-5.2: 74.7 (#89)

BenchmarkDeepSeek-R1GPT-5.2
LMArena Instruction Following13821417
LiveBench Instruction Following80.5%—
IFEval78.4%—

Long Context DeepSeek-R1 leads

DeepSeek-R1: 45.4 (#36), GPT-5.2: 44.0 (#78)

BenchmarkDeepSeek-R1GPT-5.2
LMArena Longer Query13911428
Fiction.LiveBench75%—
CL-bench—18.2%

Writing & Preference GPT-5.2 leads

DeepSeek-R1: 61.4 (#88), GPT-5.2: 66.8 (#32)

BenchmarkDeepSeek-R1GPT-5.2
LMArena Text14281439
LMArena Creative Writing14051401
EQ-Bench Creative Writing15001703
LMArena Multi-Turn14051458
Short-Story Creative Writing83%—
WildBench82.8%—
LiveBench Language48.5%—

Frequently asked questions

Is DeepSeek-R1 better than GPT-5.2?

GPT-5.2 is the stronger model overall, scoring 54.1 to 42.3 on the Noometry Index. DeepSeek-R1 costs 5.3× less per token, which makes it the better buy when GPT-5.2's lead doesn't matter for your workload.

Which is cheaper, DeepSeek-R1 or GPT-5.2?

DeepSeek-R1 is cheaper. It lists at $0.50 per million input tokens and $2.15 per million output tokens; GPT-5.2 lists at $1.75 and $14.

Is DeepSeek-R1 or GPT-5.2 better for coding?

GPT-5.2 scores higher on coding benchmarks: 51.6 versus 46.3 in the Noometry coding category.

Which has the bigger context window?

GPT-5.2 does, with 400K tokens against 164K.

How many benchmarks do DeepSeek-R1 and GPT-5.2 share?

32 benchmarks have published results for both models. DeepSeek-R1 has 52 scored results on Noometry and GPT-5.2 has 67.

Related comparisons

Go deeper