# DeepSeek-R1 vs Gemini 2.5 Flash

> DeepSeek-R1 is the stronger model overall, scoring 42.3 to 39.3 on the Noometry Index.

- Canonical page: https://noometry.com/compare/deepseek-r1-vs-gemini-2-5-flash
- Last updated: 2026-10-11
- Shared benchmarks: 39

## Summary

- They share 39 benchmarks with published results for both. DeepSeek-R1 scores higher in 6 categories and Gemini 2.5 Flash in 3 categories; 6 gaps are clear of the uncertainty.
- The widest gap is in coding, where DeepSeek-R1 leads 46.3 to 35.8.
- The biggest single-benchmark swing is GPQA (HELM): 66.6% for DeepSeek-R1 and 39% for Gemini 2.5 Flash.
- Gemini 2.5 Flash is cheaper at $0.30 / $2.50 per million input/output tokens, against $0.50 / $2.15 for DeepSeek-R1.
- Gemini 2.5 Flash accepts more context: 1.05M tokens versus 164K.

## Snapshot

| | DeepSeek-R1 | Gemini 2.5 Flash |
|---|---|---|
| Provider | DeepSeek | Google |
| Noometry Index | 42.3 | 39.3 |
| Rank | 115 | 170 |
| Context | 164K | 1.05M |
| Input $/M | $0.50 | $0.30 |
| Output $/M | $2.15 | $2.50 |
| Weights | Proprietary | Proprietary |

## Coding

- DeepSeek-R1: 46.3 (#68)
- Gemini 2.5 Flash: 35.8 (#220)

| Benchmark | DeepSeek-R1 | Gemini 2.5 Flash |
|---|---|---|
| Aider Polyglot | 71.4% | 55.1% |
| WeirdML | 41.6% | 41.9% |
| LMArena Coding | 1427 | 1424 |
| ALE-Bench | 804.12 | 661.88 |
| SWE-bench Verified (bash only) | — | 28.7% |
| SciCode | 35.7% | — |
| LiveBench Coding | 66.7% | — |
| AlgoTune | 1.7 | — |

## Agentic & Tool Use

- DeepSeek-R1: 30.7 (#75)
- Gemini 2.5 Flash: 30.8 (#74)

| Benchmark | DeepSeek-R1 | Gemini 2.5 Flash |
|---|---|---|
| BALROG | 34.9% | 33.5% |
| Terminal-Bench | — | 17.1% |
| Berkeley Function Calling Leaderboard | — | 56.2% |
| TheAgentCompany | — | 41.1% |
| DeepResearch Bench | 35.1% | — |
| METR Time Horizons | 53.8% | — |
| Vending-Bench 2 | — | 548.84 |

## Reasoning

- DeepSeek-R1: 18.6 (#278)
- Gemini 2.5 Flash: 18.1 (#286)

| Benchmark | DeepSeek-R1 | Gemini 2.5 Flash |
|---|---|---|
| ARC-AGI-2 | 1.3% | 2.5% |
| SimpleBench | 40.8% | 41.2% |
| Kagi LLM Benchmark | 69.4% | 56.8% |
| ARC-AGI-1 | 21.2% | 33.3% |
| CritPt | 1.1% | 1.1% |
| LMArena Hard Prompts | 1416 | 1422 |
| Epoch Capabilities Index | 141.29 | 143.03 |
| ForecastBench | 60 | 60.6 |
| EnigmaEval | — | 2.7% |
| LiveBench Reasoning | 83.2% | — |
| DTBench | — | 76.5% |
| LiveBench Data Analysis | 69.8% | — |
| LMCA | — | 27.5% |
| LiveBench | 71.6% | — |

## Math

- DeepSeek-R1: 43.8 (#79)
- Gemini 2.5 Flash: 39.9 (#98)

| Benchmark | DeepSeek-R1 | Gemini 2.5 Flash |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 66.4% | 73.1% |
| Omni-MATH | 42.4% | 38.5% |
| LMArena Math | 1400 | 1415 |
| LiveBench Math | 80.7% | — |
| MATH Level 5 | 96.6% | — |
| FrontierMath (Feb 2025 set) | — | 4.8% |
| FrontierMath Tier 4 (v1) | — | 4.2% |

## Knowledge

- DeepSeek-R1: 44.5 (#87)
- Gemini 2.5 Flash: 36.4 (#168)

| Benchmark | DeepSeek-R1 | Gemini 2.5 Flash |
|---|---|---|
| MMLU-Pro | 79.3% | 63.9% |
| Confabulations | 12.7% | 16.8% |
| Vectara Hallucination Rate | 11.3% | 7.8% |
| GPQA (HELM) | 66.6% | 39% |
| LMArena Expert | 1394 | 1426 |
| GPQA Diamond | 76.3% | — |
| Humanity's Last Exam | — | 12.1% |

## Multimodal

- DeepSeek-R1: —
- Gemini 2.5 Flash: 41.8 (#32)

| Benchmark | DeepSeek-R1 | Gemini 2.5 Flash |
|---|---|---|
| LMArena Vision | — | 1253 |
| GeoBench | — | 76% |
| VPCT | — | 46.2% |
| SpatialViz-Bench | — | 36.9% |

## Multilingual

- DeepSeek-R1: 52.4 (#85)
- Gemini 2.5 Flash: 52.3 (#88)

| Benchmark | DeepSeek-R1 | Gemini 2.5 Flash |
|---|---|---|
| LMArena Non-English | 1412 | 1409 |
| LMArena Chinese | 1442 | 1450 |
| LMArena French | 1417 | 1433 |
| LMArena German | 1404 | 1418 |
| LMArena Japanese | 1391 | 1405 |
| LMArena Korean | 1360 | 1385 |
| LMArena Russian | 1423 | 1415 |
| LMArena Spanish | 1411 | 1421 |

## Instruction Following

- DeepSeek-R1: 72.0 (#143)
- Gemini 2.5 Flash: 75.7 (#54)

| Benchmark | DeepSeek-R1 | Gemini 2.5 Flash |
|---|---|---|
| IFEval | 78.4% | 89.8% |
| LMArena Instruction Following | 1382 | 1405 |
| LiveBench Instruction Following | 80.5% | — |

## Long Context

- DeepSeek-R1: 45.4 (#36)
- Gemini 2.5 Flash: 47.5 (#17)

| Benchmark | DeepSeek-R1 | Gemini 2.5 Flash |
|---|---|---|
| Fiction.LiveBench | 75% | 77.8% |
| LMArena Longer Query | 1391 | 1419 |

## Writing & Preference

- DeepSeek-R1: 61.4 (#88)
- Gemini 2.5 Flash: 53.8 (#157)

| Benchmark | DeepSeek-R1 | Gemini 2.5 Flash |
|---|---|---|
| LMArena Text | 1428 | 1417 |
| LMArena Creative Writing | 1405 | 1400 |
| Short-Story Creative Writing | 83% | 76.5% |
| EQ-Bench Creative Writing | 1500 | 1137 |
| WildBench | 82.8% | 81.7% |
| LMArena Multi-Turn | 1405 | 1408 |
| LiveBench Language | 48.5% | — |

## FAQ

### Is DeepSeek-R1 better than Gemini 2.5 Flash?

DeepSeek-R1 is the stronger model overall, scoring 42.3 to 39.3 on the Noometry Index.

### Which is cheaper, DeepSeek-R1 or Gemini 2.5 Flash?

Gemini 2.5 Flash is cheaper. It lists at $0.30 per million input tokens and $2.50 per million output tokens; DeepSeek-R1 lists at $0.50 and $2.15.

### Is DeepSeek-R1 or Gemini 2.5 Flash better for coding?

DeepSeek-R1 scores higher on coding benchmarks: 46.3 versus 35.8 in the Noometry coding category.

### Which has the bigger context window?

Gemini 2.5 Flash does, with 1.05M tokens against 164K.

### How many benchmarks do DeepSeek-R1 and Gemini 2.5 Flash share?

39 benchmarks have published results for both models. DeepSeek-R1 has 52 scored results on Noometry and Gemini 2.5 Flash has 54.