# Gemini 3.8 Flash vs GPT-5.2

> Gemini 3.8 Flash is the stronger model overall, scoring 59.5 to 54.1 on the Noometry Index.

- Canonical page: https://noometry.com/compare/gemini-3-8-flash-vs-gpt-5-2
- Last updated: 2026-10-10
- Shared benchmarks: 36

## Summary

- They share 36 benchmarks with published results for both. Gemini 3.8 Flash scores higher in 9 categories and GPT-5.2 in 1 category; 9 gaps are clear of the uncertainty.
- The widest gap is in knowledge, where Gemini 3.8 Flash leads 74.9 to 59.2.
- The biggest single-benchmark swing is ProofBench: 48% for Gemini 3.8 Flash and 15% for GPT-5.2.
- Gemini 3.8 Flash is cheaper at $0.75 / $3.75 per million input/output tokens, against $1.75 / $14 for GPT-5.2.
- Gemini 3.8 Flash accepts more context: 1.05M tokens versus 400K.

## Snapshot

| | Gemini 3.8 Flash | GPT-5.2 |
|---|---|---|
| Provider | Google | OpenAI |
| Noometry Index | 59.5 | 54.1 |
| Rank | 13 | 31 |
| Context | 1.05M | 400K |
| Input $/M | $0.75 | $1.75 |
| Output $/M | $3.75 | $14 |
| Weights | Proprietary | Proprietary |

## Coding

- Gemini 3.8 Flash: 56.4 (#17)
- GPT-5.2: 52.2 (#32)

| Benchmark | Gemini 3.8 Flash | GPT-5.2 |
|---|---|---|
| LMArena WebDev | 1581 | 1415 |
| LMArena Coding | 1512 | 1447 |
| ALE-Bench | 1,270 | 1,294 |
| SWE-bench Verified | — | 73.8% |
| DeepSWE | 73.8% | — |
| FrontierCode | 41.2% | — |
| SWE-bench Verified (bash only) | — | 72.8% |
| CursorBench | 39.6% | — |
| SWE-bench Multilingual | — | 66.7% |
| FrontierSWE | 19.6% | — |
| SciCode | 56.6% | — |
| GSO | — | 27.4% |
| WeirdML | — | 72.2% |
| AlgoTune | — | 2.05 |

## Agentic & Tool Use

- Gemini 3.8 Flash: 40.3 (#22)
- GPT-5.2: 40.0 (#24)

| Benchmark | Gemini 3.8 Flash | GPT-5.2 |
|---|---|---|
| Vending-Bench 2 | 5,094 | 3,591 |
| Terminal-Bench | — | 64.9% |
| APEX-Agents | 64.3% | — |
| 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% |
| DeepResearch Bench | — | 41.1% |
| GDP.pdf | 23.4% | — |
| LMArena Search | — | 1207 |
| METR Time Horizons | — | 75.3% |

## Reasoning

- Gemini 3.8 Flash: 64.7 (#13)
- GPT-5.2: 50.1 (#33)

| Benchmark | Gemini 3.8 Flash | GPT-5.2 |
|---|---|---|
| NYT Connections (extended) | 97.4% | 83.6% |
| Chess Puzzles | 61% | 49% |
| LMArena Hard Prompts | 1506 | 1445 |
| Mystery Game Puzzles | 47% | 23% |
| DTBench | 95.7% | 90.9% |
| LMCA | 52.9% | 43.9% |
| Epoch Capabilities Index | 157.13 | 153.45 |
| ARC-AGI-2 | — | 52.9% |
| SimpleBench | — | 45.8% |
| Kagi LLM Benchmark | — | 73.3% |
| ARC-AGI-1 | — | 86.2% |
| CritPt | 18.3% | — |
| EnigmaEval | — | 10.4% |
| EBR-Bench | — | 23% |
| Surface Evolver Bench | 76.9% | — |
| ForecastBench | — | 60.1 |

## Math

- Gemini 3.8 Flash: 65.2 (#22)
- GPT-5.2: 59.8 (#32)

| Benchmark | Gemini 3.8 Flash | GPT-5.2 |
|---|---|---|
| FrontierMath (Tiers 1-3) | 68.4% | 67.4% |
| FrontierMath Tier 4 | 22% | 31.7% |
| OTIS Mock AIME 2024-2025 | 98.9% | 96.1% |
| ProofBench | 48% | 15% |
| LMArena Math | 1528 | 1439 |
| MathArena Final-Answer Competitions | — | 72% |
| FrontierMath (Feb 2025 set) | — | 40.7% |
| FrontierMath Tier 4 (v1) | — | 18.8% |

## Knowledge

- Gemini 3.8 Flash: 74.9 (#2)
- GPT-5.2: 59.2 (#29)

| Benchmark | Gemini 3.8 Flash | GPT-5.2 |
|---|---|---|
| GPQA Diamond | 95.4% | 91.4% |
| Humanity's Last Exam | 44.5% | 27.8% |
| SimpleQA Verified | 69.7% | 37.1% |
| LMArena Expert | 1529 | 1445 |
| Vectara Hallucination Rate | — | 8.4% |

## Multimodal

- Gemini 3.8 Flash: 40.9 (#33)
- GPT-5.2: 50.9 (#7)

| Benchmark | Gemini 3.8 Flash | GPT-5.2 |
|---|---|---|
| LMArena Vision | 1312 | 1268 |
| Furniture Assembly | 31.7% | 38.3% |
| VPCT | — | 84% |
| Blueprint-Bench 2 | 38.6% | — |
| LMArena Document | — | 1405 |

## Multilingual

- Gemini 3.8 Flash: 57.9 (#5)
- GPT-5.2: 53.6 (#63)

| Benchmark | Gemini 3.8 Flash | GPT-5.2 |
|---|---|---|
| LMArena Non-English | 1486 | 1425 |
| LMArena Chinese | 1547 | 1460 |
| LMArena French | 1502 | 1456 |
| LMArena German | 1470 | 1448 |
| LMArena Japanese | 1505 | 1421 |
| LMArena Korean | 1479 | 1390 |
| LMArena Russian | 1507 | 1440 |
| LMArena Spanish | 1492 | 1433 |

## Instruction Following

- Gemini 3.8 Flash: 78.2 (#11)
- GPT-5.2: 74.9 (#83)

| Benchmark | Gemini 3.8 Flash | GPT-5.2 |
|---|---|---|
| LMArena Instruction Following | 1489 | 1417 |

## Long Context

- Gemini 3.8 Flash: 46.2 (#23)
- GPT-5.2: 44.0 (#74)

| Benchmark | Gemini 3.8 Flash | GPT-5.2 |
|---|---|---|
| LMArena Longer Query | 1506 | 1428 |
| CL-bench | — | 18.2% |

## Writing & Preference

- Gemini 3.8 Flash: 72.1 (#16)
- GPT-5.2: 67.0 (#30)

| Benchmark | Gemini 3.8 Flash | GPT-5.2 |
|---|---|---|
| LMArena Text | 1494 | 1439 |
| LMArena Creative Writing | 1485 | 1402 |
| EQ-Bench Creative Writing | 1748 | 1703 |
| LMArena Multi-Turn | 1503 | 1458 |

## FAQ

### Is Gemini 3.8 Flash better than GPT-5.2?

Gemini 3.8 Flash is the stronger model overall, scoring 59.5 to 54.1 on the Noometry Index.

### Which is cheaper, Gemini 3.8 Flash or GPT-5.2?

Gemini 3.8 Flash is cheaper. It lists at $0.75 per million input tokens and $3.75 per million output tokens; GPT-5.2 lists at $1.75 and $14.

### Is Gemini 3.8 Flash or GPT-5.2 better for coding?

Gemini 3.8 Flash scores higher on coding benchmarks: 56.4 versus 52.2 in the Noometry coding category.

### Which has the bigger context window?

Gemini 3.8 Flash does, with 1.05M tokens against 400K.

### How many benchmarks do Gemini 3.8 Flash and GPT-5.2 share?

36 benchmarks have published results for both models. Gemini 3.8 Flash has 46 scored results on Noometry and GPT-5.2 has 67.