LLM
GA
Proprietary
Gemini 3.8 Flash
High-volume agentic coding and long-context work at low cost
Release2026-09-02
Context1.05M
Max output65,536 tokens
ParametersUndisclosed
Local / VRAMAPI only
LicenseProprietary
Overview
Gemini 3.8 Flash is a model in the LLM category from Google DeepMind. It was released in 2026-09-02. The vendor is based in US/UK. Best for: High-volume agentic coding and long-context work at low cost.
Notes: Launched with the restricted Gemini 3.8 Flash Cyber (Fairwind Program only). Replaces Gemini 3.7 Flash, deprecated 2026-10-08.
Strengths
- Near-frontier agentic coding at Flash pricing (vendor: 73.7% DeepSWE v1.1)
- 1M context
- Low / medium / high thinking levels
Limits
- Introductory price doubles on 2027-01-01
- Text output only
Specifications
| Vendor | Google DeepMind |
|---|---|
| Category | LLM |
| Release | 2026-09-02 |
| Status | GA |
| License | Proprietary |
| Commercial use | Not specified |
| Context window | 1.05M |
| Max output tokens | 65,536 |
| Input modalities | Text, Image, Video, Audio, PDF |
| Output modalities | Text |
| Prompt types | Text |
| Negative prompts | No |
| Max outputs | 1 |
| Local deployment | No |
| API access | Yes |
| Pricing | $0.75 / $3.75 per 1M token (intro through 2026-12-31); $1.50 / $7.50 from 2027-01-01 |
| Confidence | High |
| Vendor country | US/UK |
| Vendor type | Frontier lab |
Pricing & hardware
$0.75 / $3.75 per 1M token (intro through 2026-12-31); $1.50 / $7.50 from 2027-01-01
API-only: local GPU memory is not applicable. API access is available.