# Gemini 3.5 Flash vs Phi-4

> Gemini 3.5 Flash is the stronger model overall, scoring 54.2 to 31.2 on the Noometry Index. Phi-4 costs 39× less per token, which makes it the better buy when Gemini 3.5 Flash's lead doesn't matter for your workload.

- Canonical page: https://noometry.com/compare/gemini-3-5-flash-vs-phi-4
- Last updated: 2026-10-11
- Shared benchmarks: 21

## Summary

- They share 21 benchmarks with published results for both. Gemini 3.5 Flash scores higher in 9 categories and Phi-4 in 0 categories; 9 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where Gemini 3.5 Flash leads 62.8 to 17.7.
- The biggest single-benchmark swing is OTIS Mock AIME 2024-2025: 95.6% for Gemini 3.5 Flash and 13.8% for Phi-4.
- Phi-4 is cheaper at $0.07 / $0.14 per million input/output tokens, against $1.50 / $9 for Gemini 3.5 Flash.
- Gemini 3.5 Flash accepts more context: 1.05M tokens versus 128K.
- Phi-4 has downloadable open weights; the other is API-only.

## Snapshot

| | Gemini 3.5 Flash | Phi-4 |
|---|---|---|
| Provider | Google | Microsoft |
| Noometry Index | 54.2 | 31.2 |
| Rank | 32 | 279 |
| Context | 1.05M | 128K |
| Input $/M | $1.50 | $0.07 |
| Output $/M | $9 | $0.14 |
| Weights | Proprietary | Open |

## Coding

- Gemini 3.5 Flash: 49.4 (#49)
- Phi-4: 34.4 (#239)

| Benchmark | Gemini 3.5 Flash | Phi-4 |
|---|---|---|
| LMArena Coding | 1492 | 1231 |
| SWE-bench Verified | 79.3% | — |
| DeepSWE | 37.4% | — |
| LMArena WebDev | 1499 | — |
| SciCode | 53.1% | — |
| WeirdML | 62.6% | — |
| BigCodeBench Instruct | — | 45.5% |
| LiveBench Coding | — | 30.7% |
| BigCodeBench Complete | — | 55.4% |
| ALE-Bench | 911.02 | — |

## Agentic & Tool Use

- Gemini 3.5 Flash: 24.7 (#114)
- Phi-4: 22.8 (#128)

| Benchmark | Gemini 3.5 Flash | Phi-4 |
|---|---|---|
| APEX-Agents | 27.5% | — |
| Berkeley Function Calling Leaderboard | — | 28.8% |
| BALROG | — | 11.6% |
| GBAEval | 6.7% | — |
| GDP.pdf | 14% | — |
| Vending-Bench 2 | 5,396 | — |

## Reasoning

- Gemini 3.5 Flash: 62.8 (#18)
- Phi-4: 17.7 (#291)

| Benchmark | Gemini 3.5 Flash | Phi-4 |
|---|---|---|
| Chess Puzzles | 50% | 1% |
| LMArena Hard Prompts | 1488 | 1220 |
| Epoch Capabilities Index | 154.46 | 130.42 |
| ARC-AGI-2 | 72.1% | — |
| SimpleBench | 76.7% | — |
| NYT Connections (extended) | 92.6% | — |
| ARC-AGI-1 | 92.5% | — |
| CritPt | 13.1% | — |
| EnigmaEval | 25.4% | — |
| EBR-Bench | 4.8% | — |
| LiveBench Reasoning | — | 47.8% |
| Mystery Game Puzzles | 32% | — |
| DTBench | 94.7% | — |
| LiveBench Data Analysis | — | 45.2% |
| LMCA | 47.1% | — |
| Surface Evolver Bench | 58.1% | — |
| ForecastBench | 59 | — |
| LiveBench | — | 41.6% |

## Math

- Gemini 3.5 Flash: 60.7 (#36)
- Phi-4: 20.8 (#285)

| Benchmark | Gemini 3.5 Flash | Phi-4 |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 95.6% | 13.8% |
| LMArena Math | 1504 | 1246 |
| FrontierMath (Tiers 1-3) | 62.8% | — |
| FrontierMath Tier 4 | 26.8% | — |
| MathArena Final-Answer Competitions | 76.3% | — |
| ProofBench | 31% | — |
| LiveBench Math | — | 42% |
| MATH Level 5 | — | 64.9% |
| FrontierMath (Feb 2025 set) | 39% | — |
| FrontierMath Tier 4 (v1) | 14.6% | — |

## Knowledge

- Gemini 3.5 Flash: 66.3 (#11)
- Phi-4: 32.6 (#209)

| Benchmark | Gemini 3.5 Flash | Phi-4 |
|---|---|---|
| GPQA Diamond | 92.8% | 56.1% |
| LMArena Expert | 1495 | 1203 |
| SimpleQA Verified | 66.2% | — |
| Confabulations | — | 29.4% |
| Vectara Hallucination Rate | — | 3.7% |
| MMLU | — | 84.8% |

## Multimodal

- Gemini 3.5 Flash: 45.7 (#15)
- Phi-4: —

| Benchmark | Gemini 3.5 Flash | Phi-4 |
|---|---|---|
| LMArena Vision | 1310 | — |
| Blueprint-Bench 2 | 33.6% | — |
| LMArena Document | 1463 | — |

## Multilingual

- Gemini 3.5 Flash: 57.0 (#13)
- Phi-4: 37.2 (#237)

| Benchmark | Gemini 3.5 Flash | Phi-4 |
|---|---|---|
| LMArena Non-English | 1476 | 1197 |
| LMArena Chinese | 1526 | 1212 |
| LMArena French | 1490 | 1224 |
| LMArena German | 1492 | 1222 |
| LMArena Japanese | 1486 | 1158 |
| LMArena Korean | 1451 | 1151 |
| LMArena Russian | 1493 | 1209 |
| LMArena Spanish | 1480 | 1234 |

## Instruction Following

- Gemini 3.5 Flash: 77.0 (#30)
- Phi-4: 60.4 (#251)

| Benchmark | Gemini 3.5 Flash | Phi-4 |
|---|---|---|
| LMArena Instruction Following | 1467 | 1201 |
| LiveBench Instruction Following | — | 58.4% |

## Long Context

- Gemini 3.5 Flash: 45.4 (#38)
- Phi-4: 36.9 (#226)

| Benchmark | Gemini 3.5 Flash | Phi-4 |
|---|---|---|
| LMArena Longer Query | 1482 | 1217 |

## Writing & Preference

- Gemini 3.5 Flash: 65.5 (#47)
- Phi-4: 40.5 (#244)

| Benchmark | Gemini 3.5 Flash | Phi-4 |
|---|---|---|
| LMArena Text | 1482 | 1217 |
| LMArena Creative Writing | 1470 | 1182 |
| LMArena Multi-Turn | 1481 | 1206 |
| Short-Story Creative Writing | — | 62.6% |
| EQ-Bench 4 | 1087 | — |
| LiveBench Language | — | 25.6% |

## FAQ

### Is Gemini 3.5 Flash better than Phi-4?

Gemini 3.5 Flash is the stronger model overall, scoring 54.2 to 31.2 on the Noometry Index. Phi-4 costs 39× less per token, which makes it the better buy when Gemini 3.5 Flash's lead doesn't matter for your workload.

### Which is cheaper, Gemini 3.5 Flash or Phi-4?

Phi-4 is cheaper. It lists at $0.07 per million input tokens and $0.14 per million output tokens; Gemini 3.5 Flash lists at $1.50 and $9.

### Is Gemini 3.5 Flash or Phi-4 better for coding?

Gemini 3.5 Flash scores higher on coding benchmarks: 49.4 versus 34.4 in the Noometry coding category.

### Which has the bigger context window?

Gemini 3.5 Flash does, with 1.05M tokens against 128K.

### How many benchmarks do Gemini 3.5 Flash and Phi-4 share?

21 benchmarks have published results for both models. Gemini 3.5 Flash has 54 scored results on Noometry and Phi-4 has 37.