# DeepSeek-V3.2-Exp vs GPT-5.2

> GPT-5.2 is the stronger model overall, scoring 54.1 to 44.9 on the Noometry Index. DeepSeek-V3.2-Exp costs 17× less per token, which makes it the better buy when GPT-5.2's lead doesn't matter for your workload.

- Canonical page: https://noometry.com/compare/deepseek-v3-2-exp-vs-gpt-5-2
- Last updated: 2026-10-10
- Shared benchmarks: 41

## Summary

- They share 41 benchmarks with published results for both. DeepSeek-V3.2-Exp 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 50.1 to 22.0.
- The biggest single-benchmark swing is ARC-AGI-2: 4% for DeepSeek-V3.2-Exp and 52.9% for GPT-5.2.
- DeepSeek-V3.2-Exp is cheaper at $0.26 / $0.38 per million input/output tokens, against $1.75 / $14 for GPT-5.2.
- GPT-5.2 accepts more context: 400K tokens versus 164K.
- DeepSeek-V3.2-Exp has downloadable open weights; the other is API-only.

## Snapshot

| | DeepSeek-V3.2-Exp | GPT-5.2 |
|---|---|---|
| Provider | DeepSeek | OpenAI |
| Noometry Index | 44.9 | 54.1 |
| Rank | 68 | 31 |
| Context | 164K | 400K |
| Input $/M | $0.26 | $1.75 |
| Output $/M | $0.38 | $14 |
| Weights | Open | Proprietary |

## Coding

- DeepSeek-V3.2-Exp: 46.4 (#61)
- GPT-5.2: 52.2 (#32)

| Benchmark | DeepSeek-V3.2-Exp | GPT-5.2 |
|---|---|---|
| SWE-bench Verified (bash only) | 70% | 72.8% |
| LMArena WebDev | 1362 | 1415 |
| SWE-bench Multilingual | 59% | 66.7% |
| WeirdML | 39.5% | 72.2% |
| LMArena Coding | 1454 | 1447 |
| SWE-bench Verified | — | 73.8% |
| Aider Polyglot | 74.2% | — |
| SciCode | 38.9% | — |
| GSO | — | 27.4% |
| ALE-Bench | — | 1,294 |
| AlgoTune | — | 2.05 |

## Agentic & Tool Use

- DeepSeek-V3.2-Exp: 37.9 (#31)
- GPT-5.2: 40.0 (#24)

| Benchmark | DeepSeek-V3.2-Exp | GPT-5.2 |
|---|---|---|
| Terminal-Bench | 39.6% | 64.9% |
| Berkeley Function Calling Leaderboard | 56.7% | 55.9% |
| Vending-Bench 2 | 1,034 | 3,591 |
| GDPval | — | 49.7% |
| Remote Labor Index | — | 2.5% |
| TheAgentCompany | 42.9% | — |
| τ²-bench Airline | — | 83% |
| τ²-bench Banking | — | 32.2% |
| τ²-bench Retail | — | 81.6% |
| τ²-bench Telecom | — | 89.7% |
| DeepResearch Bench | — | 41.1% |
| LMArena Search | — | 1207 |
| METR Time Horizons | — | 75.3% |

## Reasoning

- DeepSeek-V3.2-Exp: 22.0 (#185)
- GPT-5.2: 50.1 (#33)

| Benchmark | DeepSeek-V3.2-Exp | GPT-5.2 |
|---|---|---|
| ARC-AGI-2 | 4% | 52.9% |
| Kagi LLM Benchmark | 52.2% | 73.3% |
| NYT Connections (extended) | 36.7% | 83.6% |
| ARC-AGI-1 | 57% | 86.2% |
| Chess Puzzles | 14% | 49% |
| LMArena Hard Prompts | 1433 | 1445 |
| DTBench | 87.7% | 90.9% |
| LMCA | 29.1% | 43.9% |
| Epoch Capabilities Index | 146.28 | 153.45 |
| SimpleBench | — | 45.8% |
| CritPt | 2.9% | — |
| EnigmaEval | — | 10.4% |
| Thematic Generalization | 65% | — |
| EBR-Bench | — | 23% |
| Mystery Game Puzzles | — | 23% |
| ForecastBench | — | 60.1 |

## Math

- DeepSeek-V3.2-Exp: 40.8 (#75)
- GPT-5.2: 59.8 (#32)

| Benchmark | DeepSeek-V3.2-Exp | GPT-5.2 |
|---|---|---|
| MathArena Final-Answer Competitions | 57.7% | 72% |
| OTIS Mock AIME 2024-2025 | 87.8% | 96.1% |
| ProofBench | 8% | 15% |
| LMArena Math | 1435 | 1439 |
| FrontierMath (Feb 2025 set) | 22.1% | 40.7% |
| FrontierMath Tier 4 (v1) | 2.1% | 18.8% |
| FrontierMath (Tiers 1-3) | — | 67.4% |
| FrontierMath Tier 4 | — | 31.7% |

## Knowledge

- DeepSeek-V3.2-Exp: 51.6 (#61)
- GPT-5.2: 59.2 (#29)

| Benchmark | DeepSeek-V3.2-Exp | GPT-5.2 |
|---|---|---|
| GPQA Diamond | 83.4% | 91.4% |
| Vectara Hallucination Rate | 5.3% | 8.4% |
| LMArena Expert | 1436 | 1445 |
| Humanity's Last Exam | — | 27.8% |
| SimpleQA Verified | — | 37.1% |

## Multimodal

- DeepSeek-V3.2-Exp: —
- GPT-5.2: 50.9 (#7)

| Benchmark | DeepSeek-V3.2-Exp | GPT-5.2 |
|---|---|---|
| LMArena Vision | — | 1268 |
| VPCT | — | 84% |
| Furniture Assembly | — | 38.3% |
| LMArena Document | — | 1405 |

## Multilingual

- DeepSeek-V3.2-Exp: 52.4 (#85)
- GPT-5.2: 53.6 (#63)

| Benchmark | DeepSeek-V3.2-Exp | GPT-5.2 |
|---|---|---|
| LMArena Non-English | 1409 | 1425 |
| LMArena Chinese | 1461 | 1460 |
| LMArena French | 1435 | 1456 |
| LMArena German | 1440 | 1448 |
| LMArena Japanese | 1377 | 1421 |
| LMArena Korean | 1372 | 1390 |
| LMArena Russian | 1423 | 1440 |
| LMArena Spanish | 1441 | 1433 |

## Instruction Following

- DeepSeek-V3.2-Exp: 74.7 (#89)
- GPT-5.2: 74.9 (#83)

| Benchmark | DeepSeek-V3.2-Exp | GPT-5.2 |
|---|---|---|
| LMArena Instruction Following | 1413 | 1417 |

## Long Context

- DeepSeek-V3.2-Exp: 47.6 (#15)
- GPT-5.2: 44.0 (#74)

| Benchmark | DeepSeek-V3.2-Exp | GPT-5.2 |
|---|---|---|
| CL-bench | 13.2% | 18.2% |
| LMArena Longer Query | 1428 | 1428 |
| Fiction.LiveBench | 83.3% | — |
| CL-bench Life | 9.5% | — |

## Writing & Preference

- DeepSeek-V3.2-Exp: 62.5 (#73)
- GPT-5.2: 67.0 (#30)

| Benchmark | DeepSeek-V3.2-Exp | GPT-5.2 |
|---|---|---|
| LMArena Text | 1425 | 1439 |
| LMArena Creative Writing | 1403 | 1402 |
| EQ-Bench Creative Writing | 1515 | 1703 |
| LMArena Multi-Turn | 1427 | 1458 |

## FAQ

### Is DeepSeek-V3.2-Exp better than GPT-5.2?

GPT-5.2 is the stronger model overall, scoring 54.1 to 44.9 on the Noometry Index. DeepSeek-V3.2-Exp costs 17× 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-V3.2-Exp or GPT-5.2?

DeepSeek-V3.2-Exp is cheaper. It lists at $0.26 per million input tokens and $0.38 per million output tokens; GPT-5.2 lists at $1.75 and $14.

### Is DeepSeek-V3.2-Exp or GPT-5.2 better for coding?

GPT-5.2 scores higher on coding benchmarks: 52.2 versus 46.4 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-V3.2-Exp and GPT-5.2 share?

41 benchmarks have published results for both models. DeepSeek-V3.2-Exp has 48 scored results on Noometry and GPT-5.2 has 67.