Sulat.com
AI models
Google logo

Model details

Gemini 3.8 Flash

Gemini 3.8 Flash is the newest entry in Google's fast-moving Flash family, building on its predecessor with training adjustments drawn from cybersecurity-domain data and long-running agentic loops. Google describes it as the strongest reasoning and coding Flash it has shipped, and independent coverage notes it is the third Flash-tier release in about six weeks, reflecting a deliberate emphasis on rapid iteration rather than waiting for a larger frontier sibling. The model retains the Flash line's compact, cost-focused positioning, offering the same speed and cost as the prior generation while pushing capability forward in software engineering tasks.

The standout gains are concentrated in long-horizon software engineering, where the model tops the DeepSWE v1.1 leaderboard at a fraction of the cost reported for larger competitors, even though most general benchmarks improve only modestly over the previous Flash. A paired cybersecurity-tuned sibling, Gemini 3.8 Flash Cyber, is being rolled out through a gated Fairwind Program aimed at trusted defenders and reportedly sits on the CWE-Bench patching Pareto frontier. Practically, this makes the standard variant a strong fit for coding assistants and multi-step agent pipelines that benefit from a wide context window and tool use, while organizations needing security analysis can explore the gated Cyber variant.

Googlegemini-3.8-flashgemini-flash

Quick Info

Powered by
Provider
Google
Model key
gemini-3.8-flash
Release date
Sep 2, 2026
Last updated
Sep 2, 2026
Input modalities
Output modalities
Capabilities

Cost

Input token cost
$0.75
Output token cost
$3.75

Limits

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

Transparent token rates

Compare Gemini 3.8 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 3.8 Flash

Google

Official sourceAnnouncement

Google's blog announcement on September 2, 2026 introduces Gemini 3.8 Flash alongside a cybersecurity-tuned sibling, Gemini 3.8 Flash Cyber, marking the company's third Flash-tier release in six weeks and framing 3.8 as its strongest reasoning and coding model yet at the same speed and cost as 3.7 Flash. The standard v The post highlights coding and reasoning gains driven by cybersecurity-domain training and long-running agentic loops, claiming 3.8 Flash tops the DeepSWE v1.1 long-horizon software engineering leaderboard at a fraction of competing costs. Gemini 3.8 Flash Cyber is gated behind Google's new Fairwind Program for trusted

Google

CoverageRelease Notes

Ars Technica reported on September 2, 2026 that Google shipped Gemini 3.8 Flash as its third Flash release in six weeks, noting the absence of a frontier Gemini Pro model since early 2026 and suggesting the rapid Flash cadence reduces the likelihood of a Gemini 3.5 Pro launch. The standard Flash is described as a workh The article situates Gemini 3.8 Flash on top of the DeepSWE leaderboard for complex software engineering problems at a lower cost than larger frontier competitors, while flagging that the model only marginally improves over 3.7 Flash on most benchmarks with larger coding-specific gains. It also notes that computer-use

Google

CoverageBenchmark

DataCamp's September 2, 2026 analysis frames Gemini 3.8 Flash as Google's most intelligent Flash-tier model to date, reporting specific scores including 90.8% on Terminal-Bench 2.1 (up from 81.6% for 3.7 Flash) and 54.9% on HLE-Verified, with Humanity's Last Exam essentially flat at 45.4% versus 45.7% previously. The p The writeup adds detail on Gemini 3.8 Flash Cyber, reporting that the variant exceeds a 70% real-world vulnerability discovery rate and sits on the CWE-Bench Pareto frontier for patching, with access restricted to trusted defenders through the Fairwind Program. The article places these results alongside comparisons to

Google

Official sourceDocumentation

Google DeepMind published the official model card for Gemini 3.8 Flash in September 2026, describing it as the next iteration in the Gemini 3 family built on Gemini 3.7 Flash, with advances targeted at software engineering and agentic knowledge workflows. The card specifies a 1M-token context window, a 64K-token output The model card lists distribution channels as the Gemini app, Gemini Enterprise Agent Platform, Google AI Studio, the Gemini API, Google AI Mode, and Google Antigravity, with downstream provider availability via application programming interfaces. Safety, ethics, intended usage, and limitations sections are included in

Videos about Gemini 3.8 Flash

More models around Gemini 3.8 Flash