Gemini 3.8 Flash Review — Best Reasoning and Coding Yet

What Is Gemini 3.8 Flash?

Gemini 3.8 Flash is Google's flagship workhorse model, announced on September 2, 2026. Google calls it the best reasoning and coding model it has shipped, at the same speed and low cost as Gemini 3.7 Flash, and it marks the third Flash release in six weeks.

There are two variants sharing the same foundational intelligence. The standard model is the most intelligent workhorse, with significant gains across software engineering, agentic tasks, and critical multi-step reasoning in specialized domains. Flash Cyber is Google's most capable cybersecurity model, with frontier-level vulnerability detection and automated patching, available to trusted defenders through the company's new Fairwind Program.

Both are accelerated by long-running agentic loops that recursively evaluate and refine the underlying models.

Gemini 3.8 Flash Pricing

The API is priced at an introductory $0.75 per 1M input tokens and $3.75 per 1M output tokens, identical to 3.7 Flash. The introductory price expires on December 31, 2026; from January 1, 2027, the standard rates of $1.50 per 1M input tokens and $7.50 per 1M output tokens apply.

Consumers get the model through Google AI Pro and Ultra subscriptions across the Gemini app, AI Mode in Google Search, and Gemini in Google Sheets. Enterprises access it through the Gemini Enterprise agent platform, and developers can start in Google AI Studio, Android Studio, and Google Antigravity.

One cost caveat: 3.8 Flash works harder on complex tasks, executing extra reasoning steps and calling tools iteratively. At higher effort levels it may use more tokens, which raises the effective cost even at the same per-token price.

• $0.75 / $3.75 per 1M tokens until Dec 31, 2026
• $1.50 / $7.50 per 1M tokens from Jan 1, 2027
• Google AI Pro $19.99/mo and Ultra for consumers
• No free API tier for the full model

Benchmarks and Performance

On DeepSWE v1.1, a long-horizon software engineering benchmark, Gemini 3.8 Flash outperforms most larger and more expensive frontier models at autonomously solving complex engineering problems end to end for a fraction of the cost.

In quantitative and professional domains it also leads: it outperforms 3.7 Flash on the Vals Finance Agent V2 and Harvey's Legal Agent Benchmark, and it scores 54.9% on HLE-Verified, showing strong multi-step reasoning across STEM, humanities, and professional fields.

For compute-efficiency-first workloads, Google notes you can use lower effort levels to minimize token overhead or continue relying on Gemini 3.7 Flash, which remains fully supported.

• Leads most larger models on DeepSWE v1.1
• Top of Vals Finance Agent V2 and Harvey Legal Agent
• 54.9% on HLE-Verified
• Effort-level controls to manage token spend

The Cyber Variant

Flash Cyber is the cybersecurity variant, and it delivers frontier-level performance in autonomous vulnerability discovery and automated patching at Flash speed and cost. It is available only to trusted defenders through the Fairwind Program, which gives prioritized access to government authorities, critical infrastructure operators, and software maintainers — it is not a public consumer model.

On the CyberGym benchmark for autonomous vulnerability discovery it surpasses both 3.5 Flash Cyber and significantly larger frontier models. Google also report it exceeds 70% success on an internal discovery benchmark spanning 20 programming languages, and on CWE-Bench it reaches a pass@1 of 47.2% against a leading frontier model's 47.8%, at lower cost.

Google is already applying it internally: the Chrome Security team saw 2.6x more correct patches, Wiz measured higher recall at lower cost on its pentest benchmark, and the Cloud Vulnerability Research team used it to find a critical foundational vulnerability in under two hours.

• Fairwind Program access for trusted defenders only
• Surpasses 3.5 Flash Cyber on CyberGym
• 47.2% pass@1 on CWE-Bench
• In use on Chrome security and Google Cloud

Safety and Guardrails

The model ships with safeguards against misuse in chemical, biological, radiological, and nuclear (CBRN) domains and cyber offense, following Google's Frontier Safety Framework. The Cyber variant carries a more permissive set of cybersecurity mitigations, which is exactly why access is restricted to trusted defenders.

Both models also show a significant leap in prompt injection robustness as measured by Gray Swan, which matters for agentic workflows where external content can be weaponized against the model.

• CBRN and cyber-offense safeguards built in
• Cyber variant gated via Fairwind Program
• Improved prompt injection robustness (Gray Swan)
• Enterprise posture via Gemini Enterprise

Alternatives

Google's own lineup is the first comparison layer. Gemini is the consumer-facing assistant across the ecosystem, Gemini Omni 1.1 Flash is Google's multimodal video generation model, and Google AI Studio is the developer playground that hosts the API. All are covered in this directory.

Outside Google, the frontier competition includes Claude by Anthropic, Grok by xAI, and ChatGPT by OpenAI, which offer their own coding and agentic strengths.

The deciding factor for most teams is cost-versus-reasoning at scale: 3.8 Flash delivers near-frontier results at its introductory Flash pricing, while the larger frontier models trade higher cost for marginal benchmark gains.

Pros and Cons

Strengths include best-in-Gemini reasoning and coding, genuinely strong low-cost agentic performance, a clear thin pricing runway before the intro rate expires, and a distinct defense-graded Cyber variant for qualifying organizations.

Trade-offs are the expiry of introductory pricing, the token-hungry behavior at high effort levels, the restricted availability of the Cyber model, and the absence of verified user ratings on this directory, since the model is only days old. The site list is also inherently subject to change as benchmarks evolve.

Pricing & Plans

Intro price $0.75 per 1M input tokens and $3.75 per 1M output tokens until December 31, 2026; then $1.50 and $7.50 from January 1, 2027. Consumers get it via Google AI Pro/Ultra; Cyber is Fairwind Program only.

Best For

Recommended use cases and scenarios where Gemini 3.8 Flash shines.

Frequently Asked Questions

Common questions about Gemini 3.8 Flash, answered.