Gemini 3.5 Flash
Gemini 3.5 Flash is a multimodal foundation model developed by Google DeepMind. It features a vast context window of 1M (one million) tokens.
파라미터
Undisclosed
컨텍스트
1M
라이선스
Proprietary
출시일
2026-06-19
API 가격
입력 가격 (1M 토큰당)
$2.7
출력 가격 (1M 토큰당)
$16.2
과금 모드: standard
강점
- ・Powerful multimodal functionality
- ・Long 1M token context processing
- ・Developed by Google DeepMind
약점
- ・Non-open-source license
- ・Lack of detailed performance metrics
- ・Closed model delivery format
활용 사례
- ・Bulk document analysis
- ・Processing of multimodal data
- ・Utilizing long context windows
심층 분석
Release Date
May 19, 2026
Announced at Google I/O 2026, generally available immediately
Input Price
$1.50/M tokens
Cached input: $0.15/M. Non-global regions: $1.65/M
Output Price
$9.00/M tokens
Non-global regions: $9.90/M
Context Window
1,048,576 input / 65,536 output
1M input tokens, 65K output tokens
Terminal-Bench 2.1
76.2%
Beats Gemini 3.1 Pro (70.3%) by +5.9pp
MCP Atlas
83.6%
Beats Gemini 3.1 Pro (78.2%) by +5.4pp
SWE-Bench Pro
55.1%
Slightly above Gemini 3.1 Pro (54.2%)
Finance Agent v2
57.9%
Massive +14.9pp improvement over 3.1 Pro (43.0%)
Speed
4x faster
Roughly 4x output tokens/second vs comparable frontier models
Knowledge Cutoff
January 2026
Per Google AI for Developers documentation
강점
- ・Beats Gemini 3.1 Pro on most coding and agentic benchmarks while running 4x faster
- ・At $1.50/$9 per 1M tokens, undercuts every other model in its capability class
- ・Frontier intelligence at Flash latency - collapses the Pro/Flash quality gap
- ・Dynamic thinking enabled by default for adaptive reasoning
- ・Excellent for sub-agent orchestration and multi-step tool use
약점
- ・Trails Gemini 3.1 Pro on Humanity's Last Exam (40.2% vs 44.4%) and ARC-AGI-2 (72.1% vs 77.1%)
- ・Higher cost than previous Flash models ($1.50 vs $0.50 input)
- ・No image generation capability
- ・Knowledge cutoff January 2026 (may miss recent events)
- ・Computer use not supported
경쟁사 비교
| Model | Arena | SWE | GPQA | Price |
|---|---|---|---|---|
| Gemini 3.5 Flash | ~1470 (est) | 55.1% (Pro) | ~88% (est) | $1.50/$9.00 per 1M |
| Gemini 3.1 Pro | ~1480 | 54.2% (Pro) | ~92% | $2.00/$12.00 per 1M |
| Claude 3.5 Sonnet | ~1460 | ~72% | ~87% | $3.00/$15.00 per 1M |
| GPT-5.2 | ~1475 | ~75% | ~90% | $5.00/$20.00 per 1M |
Gemini 3.5 Flash, released at Google I/O 2026 (May 19), is the first model in the Gemini 3.5 family. It delivers frontier-level agentic and coding intelligence at Flash speed and cost ($1.50/$9 per 1M tokens). It beats Gemini 3.1 Pro on most real-world coding and agentic benchmarks while running 4x faster, collapsing the traditional two-tier Pro/Flash quality distinction.
출처
분석 생성일: 2026-05-24