# Claude Haiku 4.5 vs Gemini 3 Flash Preview

> Gemini 3 Flash Preview is the stronger model overall, scoring 52.3 to 39.5 on the Noometry Index.

- Canonical page: https://noometry.com/compare/claude-haiku-4-5-vs-gemini-3-flash-preview
- Last updated: 2026-10-12
- Shared benchmarks: 42

## Summary

- They share 42 benchmarks with published results for both. Claude Haiku 4.5 scores higher in 0 categories and Gemini 3 Flash Preview in 10 categories; 9 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where Gemini 3 Flash Preview leads 49.2 to 15.1.
- The biggest single-benchmark swing is NYT Connections (extended): 14.3% for Claude Haiku 4.5 and 83.1% for Gemini 3 Flash Preview.
- Gemini 3 Flash Preview is cheaper at $0.50 / $3 per million input/output tokens, against $1 / $5 for Claude Haiku 4.5.
- Gemini 3 Flash Preview accepts more context: 1.05M tokens versus 200K.

## Snapshot

| | Claude Haiku 4.5 | Gemini 3 Flash Preview |
|---|---|---|
| Provider | Anthropic | Google |
| Noometry Index | 39.5 | 52.3 |
| Rank | 165 | 40 |
| Context | 200K | 1.05M |
| Input $/M | $1 | $0.50 |
| Output $/M | $5 | $3 |
| Weights | Proprietary | Proprietary |

## Coding

- Claude Haiku 4.5: 44.0 (#78)
- Gemini 3 Flash Preview: 50.9 (#42)

| Benchmark | Claude Haiku 4.5 | Gemini 3 Flash Preview |
|---|---|---|
| SWE-bench Verified (bash only) | 66.6% | 75.8% |
| LMArena WebDev | 1330 | 1439 |
| SWE-bench Multilingual | 64.7% | 72.7% |
| WeirdML | 45.4% | 61.6% |
| LMArena Coding | 1453 | 1460 |
| ALE-Bench | 653.48 | 1,367 |
| SWE-bench Verified | — | 75.4% |
| SciCode | 43.3% | — |
| GSO | — | 9.8% |

## Agentic & Tool Use

- Claude Haiku 4.5: 33.6 (#52)
- Gemini 3 Flash Preview: 38.7 (#29)

| Benchmark | Claude Haiku 4.5 | Gemini 3 Flash Preview |
|---|---|---|
| Terminal-Bench | 35.5% | 64.3% |
| DeepResearch Bench | 45.5% | 49.8% |
| BALROG | 31.2% | 48.1% |
| Vending-Bench 2 | 458.89 | 3,635 |
| Berkeley Function Calling Leaderboard | 68.7% | — |
| τ²-bench Airline | — | 82.5% |
| τ²-bench Banking | — | 27.3% |
| τ²-bench Retail | — | 76.8% |
| τ²-bench Telecom | — | 91.2% |
| ExploitBench | 13.7% | — |
| GDP.pdf | — | 10% |
| LMArena Search | — | 1198 |

## Reasoning

- Claude Haiku 4.5: 15.1 (#320)
- Gemini 3 Flash Preview: 49.2 (#37)

| Benchmark | Claude Haiku 4.5 | Gemini 3 Flash Preview |
|---|---|---|
| ARC-AGI-2 | 4% | 33.6% |
| NYT Connections (extended) | 14.3% | 83.1% |
| ARC-AGI-1 | 47.7% | 84.7% |
| Chess Puzzles | 8% | 40% |
| LMArena Hard Prompts | 1420 | 1465 |
| DTBench | 73.6% | 89.1% |
| LMCA | 30.9% | 43.1% |
| Epoch Capabilities Index | 142.41 | 151.8 |
| ForecastBench | 61.4 | 58.5 |
| SimpleBench | — | 61.1% |
| CritPt | 0% | — |
| Mystery Game Puzzles | — | 26% |

## Math

- Claude Haiku 4.5: 44.9 (#78)
- Gemini 3 Flash Preview: 51.7 (#55)

| Benchmark | Claude Haiku 4.5 | Gemini 3 Flash Preview |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 66.7% | 95.6% |
| LMArena Math | 1396 | 1473 |
| FrontierMath (Feb 2025 set) | 5.9% | 35.6% |
| FrontierMath Tier 4 (v1) | 2.1% | 4.2% |
| FrontierMath (Tiers 1-3) | — | 51.2% |
| FrontierMath Tier 4 | — | 17.1% |
| MathArena Final-Answer Competitions | — | 67.6% |
| ProofBench | — | 15% |
| Omni-MATH | 56.1% | — |
| MATH Level 5 | 96.4% | — |

## Knowledge

- Claude Haiku 4.5: 37.7 (#153)
- Gemini 3 Flash Preview: 58.8 (#33)

| Benchmark | Claude Haiku 4.5 | Gemini 3 Flash Preview |
|---|---|---|
| GPQA Diamond | 71.2% | 89.4% |
| SimpleQA Verified | 13.2% | 66.8% |
| Vectara Hallucination Rate | 9.8% | 13.5% |
| LMArena Expert | 1442 | 1462 |
| MMLU-Pro | 77.7% | — |
| GPQA (HELM) | 60.5% | — |

## Multimodal

- Claude Haiku 4.5: 26.8 (#118)
- Gemini 3 Flash Preview: 45.5 (#16)

| Benchmark | Claude Haiku 4.5 | Gemini 3 Flash Preview |
|---|---|---|
| Blueprint-Bench 2 | 0% | 0% |
| LMArena Document | 1420 | 1413 |
| LMArena Vision | — | 1285 |
| GeoBench | — | 88% |
| VPCT | — | 72.6% |

## Multilingual

- Claude Haiku 4.5: 49.9 (#129)
- Gemini 3 Flash Preview: 55.7 (#27)

| Benchmark | Claude Haiku 4.5 | Gemini 3 Flash Preview |
|---|---|---|
| LMArena Non-English | 1377 | 1458 |
| LMArena Chinese | 1417 | 1511 |
| LMArena French | 1408 | 1477 |
| LMArena German | 1375 | 1497 |
| LMArena Japanese | 1339 | 1489 |
| LMArena Korean | 1347 | 1443 |
| LMArena Russian | 1381 | 1480 |
| LMArena Spanish | 1420 | 1469 |

## Instruction Following

- Claude Haiku 4.5: 71.4 (#149)
- Gemini 3 Flash Preview: 75.7 (#56)

| Benchmark | Claude Haiku 4.5 | Gemini 3 Flash Preview |
|---|---|---|
| LMArena Instruction Following | 1414 | 1437 |
| IFEval | 80.1% | — |

## Long Context

- Claude Haiku 4.5: 43.6 (#92)
- Gemini 3 Flash Preview: 44.4 (#67)

| Benchmark | Claude Haiku 4.5 | Gemini 3 Flash Preview |
|---|---|---|
| LMArena Longer Query | 1427 | 1452 |

## Writing & Preference

- Claude Haiku 4.5: 57.9 (#123)
- Gemini 3 Flash Preview: 65.5 (#45)

| Benchmark | Claude Haiku 4.5 | Gemini 3 Flash Preview |
|---|---|---|
| LMArena Text | 1396 | 1466 |
| LMArena Creative Writing | 1372 | 1457 |
| LMArena Multi-Turn | 1409 | 1471 |
| WildBench | 83.9% | — |
| EQ-Bench 4 | 1064 | — |

## FAQ

### Is Claude Haiku 4.5 better than Gemini 3 Flash Preview?

Gemini 3 Flash Preview is the stronger model overall, scoring 52.3 to 39.5 on the Noometry Index.

### Which is cheaper, Claude Haiku 4.5 or Gemini 3 Flash Preview?

Gemini 3 Flash Preview is cheaper. It lists at $0.50 per million input tokens and $3 per million output tokens; Claude Haiku 4.5 lists at $1 and $5.

### Is Claude Haiku 4.5 or Gemini 3 Flash Preview better for coding?

Gemini 3 Flash Preview scores higher on coding benchmarks: 50.9 versus 44.0 in the Noometry coding category.

### Which has the bigger context window?

Gemini 3 Flash Preview does, with 1.05M tokens against 200K.

### How many benchmarks do Claude Haiku 4.5 and Gemini 3 Flash Preview share?

42 benchmarks have published results for both models. Claude Haiku 4.5 has 53 scored results on Noometry and Gemini 3 Flash Preview has 59.