Gemini 3.8 Flash Reaches General Availability
Google released Gemini 3.8 Flash on September 2, 2026. This update deprecates the previous version and targets complex enterprise workflows.
Release Details and Deprecation
Google confirmed the general availability of Gemini 3.8 Flash on September 2, 2026. This date marks the official release event recorded in the provider’s changelog. The announcement distinguishes this historical release date from the current publication date of this article.
The release coincides with the deprecation of Gemini 3.7 Flash. According to the official changelog, all requests directed to gemini-3.7-flash are automatically routed to gemini-3.8-flash. Developers are advised to update their model strings to reflect the new identifier.
Documented Specifications
The source facts establish specific technical parameters for the new model. Gemini 3.8 Flash accepts both text and image modalities as input. Its output modality is limited to text. The model features a context window of 1,048,576 tokens and a maximum output limit of 65,536 tokens.
The provider describes the model as engineered for long-horizon software engineering, autonomous agents, and complex enterprise workflows. The open weights status is recorded as false, indicating that open weights are not confirmed. No pricing data is provided in the source facts.
- Input Modalities: Text, Image
- Output Modality: Text
- Context Window: 1,048,576 tokens
- Max Output Tokens: 65,536
Evaluation Procedure
Engineers can evaluate the model by designing targeted tests based on its documented capabilities. A practical procedure involves testing the model’s handling of long-context inputs up to the 1,048,576-token limit. Measure the latency and accuracy of responses when processing multi-modal inputs containing both text and images.
Compare the output quality against baseline expectations for software engineering tasks. Validate that the automatic routing from the deprecated model string functions correctly in existing pipelines. These steps allow for a factual assessment of performance without assuming unverified benefits or architectural details.
Artificial Analysis benchmark review
Benchmark review updated on . These scores apply to the linked artifact and recorded methodology; they do not establish the performance of a different runtime or quantized model.
| Measurement | Recorded result | Measured |
|---|---|---|
| Artificial Analysis Coding Index (v4.1) ↗ | 76.3 index_points | 2026-09-02 |
| Artificial Analysis Coding Index (v4.2) ↗ | 76.3 index_points | 2026-09-05 |
| Artificial Analysis Coding Index (v4.3) ↗ | 76.3 index_points | 2026-09-20 |
| Artificial Analysis Intelligence Index (v4.1) ↗ | 58.7 index_points | 2026-09-02 |
| Artificial Analysis Intelligence Index (v4.2) ↗ | 47.1 index_points | 2026-09-05 |
| Artificial Analysis Intelligence Index (v4.3) ↗ | 40.9 index_points | 2026-09-20 |
Review the full benchmark provenance and source methodology →
Primary sources
Frequently asked questions
What happens to existing requests using the old model name?
Requests to gemini-3.7-flash are automatically routed to gemini-3.8-flash. Users should update their configuration to use the new model string directly.
Does the source confirm local installation options?
No. The source records that open weights are not confirmed. It does not establish whether local installation is possible or if API access is mandatory.