Gemini 3.5 Flash GA Release and Evaluation
Google released Gemini 3.5 Flash as a GA model on May 19, 2026. This article details confirmed specs, deprecation status, and evaluation steps for agentic…
Release Specifications and Status
The source event date is May 19, 2026. On this date, Google designated gemini-3.5-flash as generally available (GA). The provider describes it as their most intelligent model for sustained frontier performance on agentic and coding tasks. At release, it became the model behind the gemini-flash-latest alias.
Documented specifications include text and image input modalities and text-only output. The context window is 1,048,576 tokens, with a maximum output of 65,536 tokens. The openWeights field is recorded as false, meaning open weights are not confirmed. No pricing data is present in the source facts.
Deprecation and Current State
Official changelog entries state that gemini-3.5-flash is deprecated and replaced by gemini-3.6-flash. Requests to the older model string are automatically routed to gemini-3.6-flash. Developers are advised to update model strings to gemini-3.6-flash or gemini-3.8-flash.
This distinction separates the historical release date from current operational status. While the GA announcement occurred on May 19, 2026, the model is no longer the primary target for new configurations. Engineers must verify which version their applications target to align with current support policies.
Independent Evaluation Procedure
To assess capabilities against the stated focus on agentic and coding tasks, engineers should implement a structured testing protocol. Define representative coding problems and multi-step agentic workflows. Measure output accuracy and consistency using automated validation scripts. Compare results against existing production baselines to identify relative strengths.
Validate that the model handles the full 1,048,576-token context window without degradation. Since the source does not provide benchmark scores or performance rankings, conclusions about superiority must be derived from independent, reproducible tests conducted by the evaluating organization.
- Test: Execute standard coding benchmarks and agentic task suites.
- Measure: Record latency, token usage, and error rates.
- Compare: Contrast results with existing production baselines.
- Validate: Confirm that image input processing aligns with expected outcomes.
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) ↗ | 70.1 index_points | 2026-08-09 |
| Artificial Analysis Coding Index (v4.2) ↗ | 70.1 index_points | 2026-09-07 |
| Artificial Analysis Coding Index (v4.3) ↗ | 70.1 index_points | 2026-09-20 |
| Artificial Analysis Intelligence Index (v4.1) ↗ | 52 index_points | 2026-08-09 |
| Artificial Analysis Intelligence Index (v4.2) ↗ | 39.7 index_points | 2026-09-07 |
| Artificial Analysis Intelligence Index (v4.3) ↗ | 32.6 index_points | 2026-09-20 |
Review the full benchmark provenance and source methodology →
Primary sources
Frequently asked questions
What are the confirmed input and output modalities?
The source confirms text and image inputs. The output modality is strictly text. The context window is 1,048,576 tokens, and the maximum output is 65,536 tokens.
Is Gemini 3.5 Flash currently active for new deployments?
No. Official documentation states that gemini-3.5-flash is deprecated and replaced by gemini-3.6-flash. Requests to the old model string are routed to the new version. Developers should update configurations to use gemini-3.6-flash or gemini-3.8-flash.