Gemini 3.0 FlashvsHaiku 4.5
Across 12 shared benchmarks, Gemini 3.0 Flash leads overall: Gemini 3.0 Flash wins 11, Haiku 4.5 wins 1, with 0 ties and an average score difference of +22.66.
Gemini 3.0 Flash
Google Deep Mind · 2025-12-17 · Chat model
Haiku 4.5
Anthropic · 2025-10-15 · Multimodal model
Gemini 3.0 Flash11 wins(92%)(8%)1 winHaiku 4.5
Benchmark scores
Grouped by capability, sorted by largest gap within each. 12 shared benchmarks.
General Knowledge
Gemini 3.0 Flash 4/4| Benchmark | Gemini 3.0 Flash | Haiku 4.5 | Diff |
|---|---|---|---|
| HLE | 43.5047 / 172 | 4.30170 / 172Normal (No Tools) | +39.20 |
| ARC-AGI-2 | 33.6030 / 62 | 1.3055 / 62Normal (No Tools) | +32.30 |
| GPQA Diamond | 90.4019 / 187 | 60.50144 / 187Normal (No Tools) | +29.90 |
| LiveBench | 56.3579 / 115Normal (No Tools) | 45.33103 / 115Normal (No Tools) | +11.02 |
Claw-style Agent Evaluation
Even 2/2| Benchmark | Gemini 3.0 Flash | Haiku 4.5 | Diff |
|---|---|---|---|
| Claw Bench | 85.7015 / 29Thinking (With Tools) | 89.4011 / 29Thinking (With Tools) | -3.70 |
| Pinch Bench | 85.2016 / 37Thinking (With Tools) | 8221 / 37Thinking (With Tools) | +3.20 |
Coding and Software Engineer
Gemini 3.0 Flash 2/2| Benchmark | Gemini 3.0 Flash | Haiku 4.5 | Diff |
|---|---|---|---|
| SWE-Bench Pro - Public | 49.6042 / 54Thinking High (With Tools) | 39.4551 / 54Extended (with tools) | +10.15 |
| SWE-bench Verified | 68.7066 / 112 | 60.6080 / 112Normal (With Tools) | +8.10 |
Math and Reasoning
Gemini 3.0 Flash 2/2| Benchmark | Gemini 3.0 Flash | Haiku 4.5 | Diff |
|---|---|---|---|
| AIME2025 | 99.708 / 107 | 3995 / 107Normal (No Tools) | +60.70 |
| FrontierMath - Tier 4 | 4.2040 / 80Normal (No Tools) | 2.1056 / 80Thinking (No Tools, 32K Budget) | +2.10 |
Agent Level Benchmark
Gemini 3.0 Flash 1/1| Benchmark | Gemini 3.0 Flash | Haiku 4.5 | Diff |
|---|---|---|---|
| τ²-Bench | 90.203 / 43 | 3342 / 43Normal (With Tools) | +57.20 |
AI Agent - Tool Usage
Gemini 3.0 Flash 1/1| Benchmark | Gemini 3.0 Flash | Haiku 4.5 | Diff |
|---|---|---|---|
| MCP-Atlas | 6220 / 27Normal (With Tools) | 40.2027 / 27Normal (With Tools) | +21.80 |
Specs
| Field | Gemini 3.0 Flash | Haiku 4.5 |
|---|---|---|
| Publisher | Google Deep Mind | Anthropic |
| Release date | 2025-12-17 | 2025-10-15 |
| Model type | Chat model | Multimodal model |
| Architecture | Dense | Dense |
| Parameters | Not available | Not available |
| Context length | 2000K | 200K |
| Max output | 64K | 64K |
API pricing
Prices use DataLearner records when available; missing fields are not inferred.
| Item | Gemini 3.0 Flash | Haiku 4.5 |
|---|---|---|
| Text input | $0.5 / 1M tokens | $1 / 1M tokens |
| Text output | $3 / 1M tokens | $5 / 1M tokens |
| Cache read | Not public | $0.1 / 1M tokens |
| Cache write | Not public | $1.25 / 1M tokens |
Summary
- Gemini 3.0 Flashleads in:General Knowledge (4/4), Coding and Software Engineer (2/2), Math and Reasoning (2/2), Agent Level Benchmark (1/1), AI Agent - Tool Usage (1/1)
- Tied in:Claw-style Agent Evaluation
On average across the 12 shared benchmarks, Gemini 3.0 Flash scores 22.66 higher.
Largest single-benchmark gap: AIME2025 — Gemini 3.0 Flash 99.70 vs Haiku 4.5 39 (+60.70).
Page generated from structured model, pricing and benchmark records. No real-time LLM is used to write the prose.