Gemini 3.8 Flash: Pricing, Thinking Levels and Output Limits

What Gemini 3.8 Flash is

Gemini 3.8 Flash was released on 2 September 2026 as Google's most capable Flash-tier model, aimed at long-horizon software engineering, autonomous agents and enterprise workflows. On LMArena's text leaderboard Gemini 3.8 Flash has 1496 Elo (rank 10, high effort); by category, 1530 Elo in Coding (rank 22, high effort), 1524 Elo in Math (rank 4, high effort), 1527 Elo in Expert (rank 16, high effort) and 1487 Elo in Creative Writing (rank 10, high effort). It carries a 1 million token context window and returns up to 64,000 output tokens.

Pricing

Input is $0.75 per million tokens and output $3.75, and that is an introductory rate rather than the price of the model. Google's own pricing page states that the introductory rate runs to 31 December 2026, and that from 1 January 2027 the standard rates of $1.50 and $7.50 per million take effect. That is a doubling. Related costs move on the same date: context storage is $0.50 per million tokens per hour through 2026 and $1.00 thereafter. Search grounding is separate again, at 5,000 free requests per month shared across all Gemini 3.x models and $14 per thousand after that. Context caching is $0.075 per million tokens through 2026 and $0.15 from 1 January 2027.

Thinking levels, and the one that was removed

Gemini 3.8 Flash exposes tunable thinking at three levels — low, medium as the default, and high. The minimal level is not supported on this model, so a call written for an earlier Gemini that accepted minimal needs changing.

Independent figures

GPQA Diamond records 95, mock AIME 99 and SimpleQA Verified 70.

Limits

The output ceiling is 64,000 tokens. The introductory price runs to 31 December 2026; from 1 January 2027 the standard rates of $1.50 and $7.50 per million apply.

Launch details and Google’s figures

Google announced Gemini 3.8 Flash on 2 September 2026 together with Gemini 3.8 Flash Cyber. Google describe a design choice behind its results: on complex tasks the model executes extra reasoning steps and calls tools iteratively, which can use more tokens. Google report 54.9% on HLE-Verified, and state that it outperforms 3.7 Flash on Vals Finance Agent V2 and Harvey's Legal Agent Benchmark. The model identifier is gemini-3.8-flash. It is available in Google Antigravity, the Gemini API through Google AI Studio and Android Studio, Stitch, Gemini Enterprise, the Gemini app for Pro and Ultra subscribers, AI Mode in Google Search, and Google Sheets. Both models ship with safeguards against chemical, biological, radiological and nuclear misuse and cyber offense under Google's Frontier Safety Framework.

Gemini 3.8 Flash Cyber

Gemini 3.8 Flash Cyber is a cybersecurity variant restricted to trusted defenders — government authorities, critical infrastructure operators and software maintainers — through Google's Fairwind Program, with more permissive mitigations suited to defensive work. Google report a success rate above 70% on an internal vulnerability-discovery evaluation across 20 programming languages, and 47.2% on CWE-Bench patching against 47.8% for a leading frontier model. Google also state that it surpasses 3.5 Flash Cyber on CyberGym and improves on prior versions in Gray Swan prompt-injection robustness testing.

Official Resources