From The Verge · Stevie Bonifield · · 1 min
Google’s Gemini 3.8 Flash Trades Tokens for Power
Google's latest AI model uses extra reasoning steps to deliver top-tier coding and agentic performance.
In brief
Google's latest AI model uses extra reasoning steps to deliver top-tier coding and agentic performance. Gemini 3.8 Flash keeps per-token rates identical to its predecessor, but total task costs can rise ~40% because the model thinks longer to produce better results. Originally reported by The Verge.
Google launches Gemini 3.8 Flash
Arriving just weeks after 3.7 Flash, Google's new model is built to work harder by running more reasoning steps and calling tools iteratively on complex tasks.
“"works harder" than Gemini 3.7 Flash by performing more reasoning steps on complex tasks”
Same rates, potentially higher bills
Per-token pricing remains unchanged, but because the model uses around 30% more output tokens per task to maximize performance, overall task costs can increase by roughly 40%.
“the model might use more tokens to maximize performance, especially at higher effort levels.”
Strong early reviews for speed and coding
Early testers praise its efficiency, comparing its output to high-end models like Anthropic's Opus 5 while delivering faster responses at a fraction of the cost.
“Opus 5 coding quality but at a fraction of the cost and super fast”
Dominating specialized AI benchmarks
The model beats competitors on software engineering tests like DeepSWE v1.1, as well as specialized legal and financial agent evaluations, while adding strict safety safeguards.
Cyber variant targets infrastructure security
Google also unveiled Gemini 3.8 Flash Cyber alongside its new Fairwind Program, allowing select government and cybersecurity partners to autonomously patch software vulnerabilities.
The bottom line
Gemini 3.8 Flash keeps per-token rates identical to its predecessor, but total task costs can rise ~40% because the model thinks longer to produce better results.
