gemini-3.5-flash

Gemini-3.5-Flash sits within Google's Gemini Flash family, the lineup the company tunes for fast, low-latency inference in production agents and assistants. Google's documentation for the 3.5 generation frames these Flash-tier releases as their most cost-effective class, prioritizing high throughput and quick response while still handling varied task complexity. The model is multimodal on input, accepting text, images, and audio, while emitting text-only outputs, making it a natural fit for chat interfaces and tool-calling pipelines that need to read rich attachments and reply quickly. Source: https://deepmind.google/models/gemini/flash-lite/

Within that family positioning, Gemini-3.5-Flash is designed for agentic and interactive workloads that benefit from the Flash tier's blend of reasoning ability, structured output support, and tooling. It is offered with a large context window and a generous output budget, suitable for multi-turn reasoning, retrieval-augmented chats, and assistants that call functions or generate JSON. The Flash line as a whole emphasizes speed and cost efficiency per token, so this version is best matched to high-volume applications where latency and predictable behavior matter more than the deepest reasoning depth of larger Gemini variants. Source: https://deepmind.google/models/gemini/flash-lite/

SAP AI Coregemini-3.5-flashgemini-flash

Open provider page Provider docs

Quick Info

Provider: SAP AI Core
Model key: gemini-3.5-flash
Release date: May 19, 2026
Last updated: May 19, 2026
Knowledge cutoff: 2025-01
Input modalities
Output modalities
Capabilities

Cost

Input token cost: $1.50
Output token cost: $9.00

Limits

Output tokens: 65,536 tokens
Context window: 1,048,576 tokens

Transparent token rates

Compare gemini-3.5-flash pricing

Rates are shown per one million tokens. Combined means one million input plus one million output tokens.

Browse this family

Latest news about gemini-3.5-flash

SAP AI Core

Coverage

Google Cloud's I/O 2026 wrap-up names Gemini 3.5 as Google's latest family of models "combining frontier intelligence with action," with 3.5 Flash as the first release. CEO Thomas Kurian's post explicitly states that 3.5 Flash "delivers frontier performance for agents and coding, excelling at complex long-horizon tasks According to the Cloud blog, Google DeepMind engineered the Gemini 3.5 models using purpose-built hardware (Google's eighth-generation TPUs), and the co-design enables deeper reasoning capabilities trained faster and more efficiently per generation. Availability on Cloud-side surfaces includes Gemini Enterprise and Wor

SAP AI Core

Coverage

Google DeepMind announced Gemini 3.5 on May 19, 2026, kicking off the family with 3.5 Flash (model ID gemini-3.5-flash), available the same day via the Gemini app, AI Mode in Google Search, Google Antigravity, the Gemini API in AI Studio and Android Studio, and Gemini Enterprise/Enterprise Agent Platform. Per the Googl The same post frames 3.5 Flash as Google's strongest agentic and coding model yet, optimized for long-horizon tasks and landing in the top-right quadrant of the Artificial Analysis index. Availability spans consumer surfaces (Gemini app, Search AI Mode) and developer/enterprise surfaces (Antigravity, Gemini API, Gemini

SAP AI Core

CoverageBenchmark

The benchmarks.company aggregator page for Gemini 3.5 Flash lists core developer-facing metadata: developer listed as Google, type Proprietary, release date 2026-05, context window 1,000K tokens, input pricing $1.5 per million tokens, and output pricing $9 per million tokens. These specs reflect Google's published list pricing for the underlying Gemini API model and are clearly labelled by the aggregator as such, making them useful as a triangulation reference rather than as a primary source. Concrete reported benchmark scores on the aggregator include SWE-bench Verified 78.8% resolved (third-party provenance, dated 2026-07), SWE-bench Pro 53.9% resolved pass@1 (developer provenance, dated 2026-05), and Terminal-Bench 2.1 76.2% tasks resolved (developer provenance, dated 2026-05), with explicit provenance and date labels so readers can distinguish self-reported from third-party measurements. The page also surfaces comparison links against other contemporary models. Although the page references some speculative forward-dated comparator model names, the Gemini 3.5 Flash figures themselves are flagged with clear provenance and provide a useful independent cross-check of the model's coding and agentic performance claims.

SAP AI Core

Official sourceRelease Notes

Google's official "What's new in Gemini 3.5 Flash" documentation confirms that the model with ID `gemini-3.5-flash` is now generally available (GA), stable, and ready for scaled production use, positioned as the most intelligent Flash model for agentic execution, coding, and long-horizon tasks at scale. The page documents several developer-relevant behavioral changes specific to Gemini 3.5 Flash: a new default effort level of "medium," thought preservation across turns, deprecation of sampling parameters and thinking_budget in the Gemini 3.x family, strict response matching in function calling, multimodal function responses, inline instructions in function responses, and reduced unnecessary tool calls. Migration checklists are provided for users moving from Gemini 3 Flash Preview and from Gemini 2.5. The page outlines Gemini 3.5 Flash's place in the broader Gemini 3 Flash family and provides prompting best practices, limitations, and FAQ guidance for production integration. It explicitly references the model identifier `gemini-3.5-flash` and is targeted at developers adopting the Interactions API, with links to the generateContent API and to the newer Gemini 3.8 Flash announcement. This is first-party Google documentation that defines the model's GA behavior, capability changes, and migration path, making it directly relevant to developers building on the underlying model rather than on any particular serving provider.