Gemini 3.8 Flash Reaches General Availability
Gemini CLIGoogle has taken Gemini 3.8 Flash generally available in the Gemini API, its third Flash-tier release in six weeks and the direct successor to Gemini 3.7 Flash. The model is built for long-horizon software engineering and autonomous agent workloads, posting a jump to 90.8% on Terminal-Bench 2.1 (up from 81.6% on 3.7 Flash) and outperforming several larger frontier models on the DeepSWE v1.1 coding benchmark. It keeps the same $0.75 per million input / $3.75 per million output introductory pricing as 3.7 Flash through the end of 2026, doubling to standard rates on January 1, 2027. Google also introduced a restricted Gemini 3.8 Flash Cyber variant for vulnerability research, gated behind its Fairwind Program and not broadly developer-accessible.
Key Takeaways
- Gemini 3.8 Flash is Google's third Flash-tier release in six weeks, following 3.6 Flash (July 21) and 3.7 Flash (August 13), an unusually rapid iteration cadence for a flagship model line.
- Terminal-Bench 2.1 score jumped to 90.8% from 3.7 Flash's 81.6%, one of the largest single-release benchmark gains in the Flash line's recent history.
- On DeepSWE v1.1, 3.8 Flash reportedly outperforms several larger, more expensive frontier models on long-horizon software engineering tasks, at a fraction of their inference cost.
- Introductory pricing stays flat at $0.75/$3.75 per million input/output tokens through the end of 2026, meaning existing 3.7 Flash workloads can migrate without a cost increase before the January 2027 rate change.
- The model's 1M-token context window and multimodal input (text, image, video, audio, PDF) make it a drop-in upgrade for existing agentic pipelines built on earlier Flash models.
- The companion Gemini 3.8 Flash Cyber variant is gated behind Google's Fairwind Program for vetted security teams, so most developers will only ever interact with the base 3.8 Flash model, not the security-tuned sibling.
Sources & Mentions
5 external resources covering this update
Gemini 3.8 Flash and 3.8 Flash Cyber
Hacker News
Google's Gemini 3.8 Flash is built for agents, while its Cyber twin hunts vulnerabilities
VentureBeat
Gemini 3.8 Flash rolling out three weeks after last release
9to5Google
Google DeepMind Releases Gemini 3.8 Flash and Gemini 3.8 Flash Cyber: One Core Model, Two Access Envelopes
MarkTechPost
Gemini 3.8 Flash: Features, Benchmarks, and Pricing
DataCamp
A New Flagship Flash Model
Google has shipped Gemini 3.8 Flash (gemini-3.8-flash) as a generally available model in the Gemini API, describing it as the company's "most intelligent Flash model, engineered for long-horizon software engineering, autonomous agents, and complex enterprise workflows." It arrives just three weeks after Gemini 3.7 Flash, making it Google's third Flash-tier launch in roughly six weeks, a pace that reflects how central the Flash line has become to Google's agentic coding push.
The model keeps Flash's now-familiar footprint: a 1,048,576-token input context window, a 65,536-token output ceiling, and support for text, image, video, audio, and PDF input with text-only output. It supports three configurable thinking levels (low, medium, and high), letting developers dial reasoning depth up for complex agentic tasks or down for everyday, latency-sensitive requests.
Performance Gains Over 3.7 Flash
Google's own benchmark comparisons show 3.8 Flash improving on every metric it publishes relative to 3.7 Flash. On Terminal-Bench 2.1, a benchmark for autonomous terminal-based task completion, the model jumps to 90.8% from 3.7 Flash's 81.6%. On DeepSWE v1.1, a long-horizon software engineering benchmark, third-party coverage reports 3.8 Flash reaching 73.7% (versus 65.3% for 3.7 Flash) and outperforming several larger frontier models at a fraction of their cost. The model achieves this by taking smaller, more deliberate reasoning steps on difficult multi-step goals, calling tools iteratively, and verifying its own work along the way, at the cost of somewhat higher token consumption on complex tasks. For everyday requests, developers can lower the reasoning effort to keep costs and latency down.
Pricing and Availability
Gemini 3.8 Flash keeps the exact same introductory pricing Google set for 3.7 Flash: $0.75 per million input tokens and $3.75 per million output tokens, in effect through December 31, 2026. Standard pricing of $1.50 / $7.50 per million tokens takes over starting January 1, 2027. The model is available now through the Gemini API, Google AI Studio, Antigravity, Android Studio, and the Gemini Enterprise Agent Platform, and it supports caching, code execution, computer use (preview), file search, function calling, Google Maps grounding, search grounding, structured outputs, URL context, the Batch API, and both Flex and Priority inference tiers.
A Restricted Security Sibling
Alongside the general model, Google introduced Gemini 3.8 Flash Cyber, a variant built on the same underlying model but tuned with more permissive cyber-mitigation behavior for vulnerability discovery and automated patching. Unlike the base 3.8 Flash model, Cyber is not broadly available: access runs through Google's Fairwind Program and is limited to vetted government authorities, critical infrastructure operators, and software maintainers. Google says its own Cloud Vulnerability Research team used the model to find a critical vulnerability in under two hours, a process that research teams typically measure in months. Because the Cyber variant sits behind an application-gated access program rather than open developer access, it functions as a specialty security tool rather than a general-purpose CLI or API upgrade.