Gemini 3.5 Flash-Lite
Google DeepMind's fastest and most cost-effective Gemini 3.5 model, released July 2026 for high-throughput execution. Flash-Lite runs at roughly 350 output tokens per second on Artificial Analysis while posting large gains over the previous Lite generation on agentic and long-context benchmarks.
Strengths
- Most cost-effective model in the Gemini 3.5 class at $0.30 per 1M input tokens
- Roughly 350 output tokens per second on Artificial Analysis
- 54% on Terminal-Bench 2.1, up from 31% for the previous generation
- 72.2% on GDM-MRCR v2 long-context retrieval, up from 60.1%
- Powers high-volume surfaces including Google Search and the Gemini app
Caveats
- Proprietary - Google API only
- Context window is not stated on the launch announcement
- Lite tier - trades peak capability for throughput and cost
- Vendor
- License
- Proprietary
- Release Date
- 2026-07-21
- Modalities
- textimageaudiovideo
| Benchmark | Score | Max |
|---|---|---|
| Terminal-Bench 2.1 | 54.0 | 100.0 |
| SWE-Bench Pro | 54.2 | 100.0 |
| GDM-MRCR v2 | 72.2 | 100.0 |
| OSWorld-Verified | 74.0 | 100.0 |
Capabilities
Vision
Audio
Video
Tool Use
Resources
Pricing
$0.30 per 1M input tokens, $2.50 per 1M output tokens - the most cost-effective model in the Gemini 3.5 class.