Gemini 3.5 FlashvsOpus 4.7
Across 11 shared benchmarks, Opus 4.7 leads overall: Gemini 3.5 Flash wins 4, Opus 4.7 wins 7, with 0 ties and an average score difference of -1.61.
Gemini 3.5 Flash
Google Deep Mind · 2026-06-20 · Multimodal model
Opus 4.7
Anthropic · 2026-04-16 · Reasoning model
Gemini 3.5 Flash4 wins(36%)(64%)7 winsOpus 4.7
Benchmark scores
Grouped by capability, sorted by largest gap within each. 11 shared benchmarks.
AI Agent - Tool Usage
Gemini 3.5 Flash 3/3| Benchmark | Gemini 3.5 Flash | Opus 4.7 | Diff |
|---|---|---|---|
| Terminal-Bench 2.1 | 76.2022 / 43Thinking High (With Tools) | 69.7031 / 43Thinking High (With Tools) | +6.50 |
| MCP-Atlas | 83.604 / 38Thinking High (With Tools) | 79.109 / 38Deep Thinking (With Tools) | +4.50 |
| OSWorld-Verified | 78.4010 / 26Thinking High (With Tools) | 7811 / 26Extended (with tools) | +0.40 |
General Knowledge
Opus 4.7 3/3| Benchmark | Gemini 3.5 Flash | Opus 4.7 | Diff |
|---|---|---|---|
| HLE | 40.2066 / 181Thinking High (With Tools) | 54.7015 / 181Extended (with tools) | -14.50 |
| ARC-AGI-2 | 72.1013 / 62Thinking High (With Tools) | 75.8011 / 62最高（无工具） | -3.70 |
| LiveBench | 75.0217 / 115Thinking High (No Tools) | 76.917 / 115Deep Thinking (No Tools) | -1.89 |
Math and Reasoning
Opus 4.7 2/2| Benchmark | Gemini 3.5 Flash | Opus 4.7 | Diff |
|---|---|---|---|
| FrontierMath v2 | 62.8118 / 34Thinking High (No Tools) | 70.1814 / 34最高（无工具） | -7.37 |
| FrontierMath Tier 4 v2 | 26.8318 / 34Thinking High (No Tools) | 31.7114 / 34最高（无工具） | -4.88 |
Coding and Software Engineer
Opus 4.7 1/1| Benchmark | Gemini 3.5 Flash | Opus 4.7 | Diff |
|---|---|---|---|
| SWE-Bench Pro - Public | 55.1030 / 57Thinking High (With Tools) | 64.308 / 57Extended (with tools) | -9.20 |
Commonsense Reasoning
Gemini 3.5 Flash 1/1| Benchmark | Gemini 3.5 Flash | Opus 4.7 | Diff |
|---|---|---|---|
| SimpleBench | 76.704 / 67Normal (No Tools) | 62.9012 / 67Normal (No Tools) | +13.80 |
General Evaluation
Opus 4.7 1/1| Benchmark | Gemini 3.5 Flash | Opus 4.7 | Diff |
|---|---|---|---|
| GPQA Diamond | 92.8019 / 226Thinking High (No Tools) | 94.205 / 226Extended (no tools) | -1.40 |
Specs
| Field | Gemini 3.5 Flash | Opus 4.7 |
|---|---|---|
| Publisher | Google Deep Mind | Anthropic |
| Release date | 2026-06-20 | 2026-04-16 |
| Model type | Multimodal model | Reasoning model |
| Architecture | Dense | Dense |
| Parameters | Not available | Not available |
| Context length | 1M | 1000K |
| Max output | 64K | 128K |
API pricing
Prices use DataLearner records when available; missing fields are not inferred.
| Item | Gemini 3.5 Flash | Opus 4.7 |
|---|---|---|
| Text input | $1.5 / 1M tokens | $5 / 1M tokens |
| Text output | $9 / 1M tokens | $25 / 1M tokens |
| Cache read | $0.15 / 1M tokens | $0.5 / 1M tokens |
| Cache write | Not public | $6.25 / 1M tokens |
Summary
- Gemini 3.5 Flashleads in:AI Agent - Tool Usage (3/3), Commonsense Reasoning (1/1)
- Opus 4.7leads in:General Knowledge (3/3), Math and Reasoning (2/2), Coding and Software Engineer (1/1), General Evaluation (1/1)
On average across the 11 shared benchmarks, Opus 4.7 scores 1.61 higher.
Largest single-benchmark gap: HLE — Gemini 3.5 Flash 40.20 vs Opus 4.7 54.70 (-14.50).
Page generated from structured model, pricing and benchmark records. No real-time LLM is used to write the prose.