Gemini 3.5 Flash
GeminiBalanced
ThinkingTool UseVisionStructured Output
About this model
Gemini 3.5 Flash is Google's high-efficiency multimodal model, bringing near-Pro level coding and reasoning at Flash-tier cost and speed. It is highly optimized for coding proficiency and parallel agentic execution...
Performance Tier
Balanced
Gemini 3.5 Flash is a balanced model from Gemini : strong performance at a reasonable price.
Strong cost-performance ratio. Reliable for most professional use cases without premium pricing.
Pricing
This model is included in Elosia plansModerate
Moderate cost. A balanced choice for regular use without constant cap watching.
| Type | per 1M tokens |
|---|---|
| Input (prompt) | $1.50 |
| Output (completion) | $9.00 |
| Image | $1.50 |
| Internal reasoning | $9.00 |
| Cache read | $0.150 |
| Cache write | $0.083 |
Capabilities
Context Length1.0M
Max Output Tokens66K
TokenizerGemini
Inputtext, image, video, file, audio
Outputtext
Release DateMay 19, 2026
Benchmarks
General Intelligence
MMLU
Not reported
Mathematics
MATH-500
Not reported
Programming
HumanEval
Not reported
Reasoning
ARC-AGI-2
72.1%
Humanity's Last Exam
40.2%
Multimodal
MMMU-Pro
83.6%
Agentic
SWE-bench Pro
55.1%
Recommended Use Cases
General ChatCodingAnalysisResearchData Extraction
Strengths
- Frontier agentic and coding performance at Flash-tier cost — ARC-AGI-2 72.1% (close to Gemini 3.1 Pro 77.1%)
- State-of-the-art multimodal understanding (MMMU-Pro 83.6%) across text, image, video, audio and PDF inputs
- Strong tool-use and computer-use workflows (Terminal-bench 2.1 76.2%, OSWorld-Verified 78.4%)
- 1M token context window with 64K tokens of output capacity
Limitations
- Preview model — behavior may change
- DeepMind reports agentic and reasoning benchmarks but not classic academic suites (MMLU, MATH-500, HumanEval, IFEval)
- Higher completion price than Gemini 3 Flash ($9/M vs lower previous Flash tiers)
Resources
This model may use your data for training
