Google DeepMind put Gemini 3.8 Flash into general availability on Sept. 2, pitching it as a coding and agent workhorse in the same race as Anthropic’s and OpenAI’s developer models.
DeepMind lists status as general availability. Gemini API docs use the stable model code gemini-3.8-flash, with about 1 million input tokens and 64,000 output tokens (1,048,576 / 65,536 in the API docs). The model takes text, image, video, audio, and PDF, and supports function calling, search as a tool, and computer use (computer use is marked preview).
Google calls it its most intelligent workhorse yet for coding and agents, the same line it used for Gemini 3.7 Flash three weeks earlier. This is the third Flash release in about six weeks.
On Google’s numbers, 3.8 Flash scores 54.9% on HLE-Verified and beats 3.7 Flash on Vals Finance Agent V2 and Harvey’s Legal Agent Benchmark. Those figures are Google’s. THE DECODER, citing Artificial Analysis, reported 59 vs 56 for 3.7 Flash on that firm’s Intelligence Index.
Pricing stays at Gemini 3.7 Flash’s introductory API rate: $0.75 per million input tokens and $3.75 per million output through Dec. 31, 2026. Standard rates of $1.50 / $7.50 start Jan. 1, 2027. Google also says the model may spend more tokens on hard tasks.
Availability spans the Gemini API, Google AI Studio, Google Antigravity, and Gemini Enterprise Agent Platform. In the Gemini app, Search AI Mode, and Sheets, Google says access is limited to AI Pro and Ultra subscribers.
A separate Gemini 3.8 Flash Cyber build is limited-access for trusted defenders through Google’s Fairwind Program and is not this general-availability coding model.