Google releases Gemini 3.8 Flash, its third Flash model in six weeks

Summary

Google has released Gemini 3.8 Flash. This is the company's third Flash-tier model released within six weeks. Google positions this model as its best-ever reasoning and coding model. However, the promised Gemini 3.5 Pro remains conspicuously absent. A specialized 'Flash Cyber' variant, optimized for security vulnerability detection and patch development, has also begun rolling out to governments and trusted testers.

Why it matters

The rapid pace of Flash releases indicates that Google is prioritizing cost-effective, high-volume models over frontier Pro versions. Since competitors have already lowered token costs, this move is reshaping the competitive pricing landscape. For developers, the top result on the DeepSWE coding benchmark and the discounted price of $0.75 / $3.75 per million tokens provide a compelling reason to shift workloads to this model. For the Chrome security team, the reported 2.6x increase in patch accuracy points to a new category of security-AI tools that could be significant beyond niche use cases.

Highlights

• Third Flash model in six weeks; no Pro version since early 2026
• Ranked #1 on the DeepSWE coding benchmark at half the regular price
• Flash Cyber: 2.6x increase in patch accuracy for the Chrome security team; critical security vulnerability found in 2 hours

Critical take

The benchmark improvements over 3.7 Flash are described as 'marginal' in most tests. Computer use performance still lags significantly behind Claude Opus. Therefore, the claim of being the 'best model to date' relies heavily on a single coding benchmark. Given that a new Flash model appears every few weeks, the entry-level price is likely temporary. The pricing emphasis remains largely cosmetic, as the discount will likely expire before the regular price takes effect.