What Is Gemini 3.8 Flash, and Should You Switch?
Google released Gemini 3.8 Flash on September 2, 2026 — its fourth Flash model in four months. It beats Claude Sonnet 5 and GPT-5.6 Sol on coding benchmarks and is already live in the Gemini app for Pro and Ultra subscribers. Here's what changed, what it costs, and whether it's worth switching.
See Claude set up for your job
Skip the theory — pick your profession and get the real workflows, ready-to-use prompts, and exact setup for your work.
Models change every month.
One short weekly update that keeps this call current — free.
If you use Google Gemini for coding, research, or multi-step workflows, the answer is yes — Gemini 3.8 Flash, released September 2, 2026, is worth switching to. It now outperforms Claude Sonnet 5 and GPT-5.6 Sol on real-world software engineering benchmarks and is already live in the Gemini app for Pro and Ultra subscribers. There is one cost caveat for API developers: despite the same per-token price as 3.7 Flash, it uses more tokens per task, making actual jobs run about 40% more expensive.
What changed
Google released Gemini 3.8 Flash on September 2, 2026 — its fourth Flash variant in four months. That release cadence is unusual, and some analysts have raised the question of whether rapid budget-model iteration is distracting Google from frontier models (Gemini 3.5 Pro and Gemini 4 remain unreleased). But the model itself is genuinely capable.
The headline improvement is coding. On DeepSWE v1.1, the leading benchmark for long-horizon software engineering tasks:
- Gemini 3.8 Flash: 73.7%
- Claude Opus 5: 74.0%
- GPT-5.6 Sol: 72.7%
- Gemini 3.7 Flash: 65.3%
- Claude Sonnet 5: 53.8%
That's a substantial jump from 3.7 Flash and a meaningful position in the competitive landscape. For coding and agentic tasks, 3.8 Flash is now among the best models available at any price tier — not just a cheaper shortcut.
Google also released Gemini 3.8 Flash Cyber, a cybersecurity-specialized variant for vulnerability discovery and automated patching. It's restricted to trusted defenders through Google's Fairwind Program and not available on general release.
Who gets access and how
Gemini app users: If you have a Google AI Pro or Ultra subscription, Gemini 3.8 Flash is already available in the Gemini app, in Google Search's AI Mode, and in Gemini in Google Sheets. You don't need to do anything — where Google has rolled it out as the default, you're already using it.
Developers: Access via Google AI Studio, Google Antigravity, and the Gemini API using model ID gemini-3.8-flash. The model supports function calling, code execution, computer use (preview), file search, search grounding, and structured outputs. Context window is 1 million tokens with 65,536 output tokens.
What it costs
Introductory pricing through December 31, 2026:
- Input: $0.75 per million tokens
- Output: $3.75 per million tokens
- Cached inputs: 90% discount
Starting January 1, 2027, pricing increases to $1.50/$7.50 per million tokens.
The per-token prices are the same as introductory pricing on 3.7 Flash, but the actual cost per completed task is roughly 40% higher. The reason: 3.8 Flash generates more intermediate steps — reasoning tokens, action steps, tool calls — to reach its answers on agentic work. The model is "working harder," as The Verge framed it, and that effort shows up in the output token count.
For Gemini Pro and Ultra subscribers using the app, this is invisible — you're not billed per token. For API developers running agents or coding pipelines, it's worth testing actual task costs before assuming the new model is cost-neutral.
What this means for the model landscape
Google's strategy with rapid Flash releases is price-performance optimization on a fast iteration loop. The pattern: release a capable, fast, cheap model; improve it aggressively within the Flash tier; use customer feedback and production deployments to close the gap with frontier models before the frontier model is ready.
The result is that "budget tier" in 2026 no longer means meaningfully worse for most professional tasks. A model that beats Claude Sonnet 5 on coding — at Flash prices — compresses the justification for paying frontier-model prices on routine coding work.
New DeepMind head Koray Kavukcuoglu has signaled that Google still wants to lead on raw capability, not just price-performance. Gemini 4 and 3.5 Pro remain on the roadmap. But for professionals making day-to-day model choices, 3.8 Flash is the practical decision point today.
What to do
- Gemini app subscribers (Pro/Ultra): Confirm you're on 3.8 Flash in Settings → Model. If it's not yet the default in your region, it's rolling out through September.
- API developers: Switch
model: "gemini-3.7-flash"tomodel: "gemini-3.8-flash"in your config. Run a cost comparison on your ten most common task types before deploying broadly. - Using Claude or GPT for coding? 3.8 Flash is now worth a side-by-side test — the benchmark gap with Sonnet 5 is large enough to affect real output quality, not just benchmark scores.
- No Gemini subscription? The model is not available on the free Gemini tier. Google AI Pro is the entry point.
Sources
- Google says its new Gemini 3.8 Flash model 'works harder' but might cost more — The Verge, September 2, 2026
- Gemini 3.8 Flash is Google's third budget model in six weeks while frontier models remain MIA — The Decoder, September 2, 2026
- Google has released Gemini 3.8 Flash, its fourth Flash model in under four months — Artificial Analysis, September 2, 2026
- Gemini 3.8 Flash rolling out three weeks after last release — 9to5Google, September 2, 2026
See Claude set up for your job
Skip the theory — pick your profession and get the real workflows, ready-to-use prompts, and exact setup for your work.
Set up AI for your job — free, in about 2 minutes
Pick your profession and get your first working AI tool, a step-by-step guide, and a $0 plugin to take home. No credit card.
Get my free setupSee Claude set up for your job
Real workflows and ready-to-use prompts, profession by profession.
Frequently asked questions
What is Gemini 3.8 Flash?+
Gemini 3.8 Flash is Google's latest budget-tier AI model, released September 2, 2026. It's the fourth Flash variant Google has released in four months and is engineered primarily for software engineering tasks, autonomous agents, and complex multi-step workflows. Despite the 'Flash' label (which historically meant fast but less capable), 3.8 Flash now matches or beats some frontier-class models on coding benchmarks.
How does Gemini 3.8 Flash compare to other AI models I use?+
On the DeepSWE v1.1 software engineering benchmark, Gemini 3.8 Flash scores 73.7%, compared to Claude Sonnet 5 at 53.8%, GPT-5.6 Sol at 72.7%, and Claude Opus 5 at 74.0% (the top performer). For coding and agentic tasks, 3.8 Flash is now genuinely competitive with the best models available — not just a cheaper shortcut. For general conversation and reasoning, the gap is narrower.
How much does Gemini 3.8 Flash cost?+
Introductory pricing through December 31, 2026: $0.75 per million input tokens and $3.75 per million output tokens. Starting January 1, 2027, pricing rises to $1.50/$7.50 per million tokens. Cached inputs receive a 90% discount. Consumer users with Gemini Pro or Ultra subscriptions get access through the Gemini app at no extra per-call charge.
Is Gemini 3.8 Flash available in the Gemini app, or only the API?+
Both. For consumers, Gemini 3.8 Flash is live now in the Gemini app for Google AI Pro and Ultra subscribers, in AI Mode in Google Search, and in Gemini in Google Sheets. Developers can access it via Google AI Studio, Google Antigravity, and the Gemini API (model ID: gemini-3.8-flash).
What's the catch — why does it cost more per task if the token prices are the same?+
Gemini 3.8 Flash uses more output tokens per task than 3.7 Flash did, especially on agentic and coding work. Artificial Analysis found that despite stable per-token pricing, the actual cost per completed task is roughly 40% higher than 3.7 Flash, because the model generates more intermediate reasoning and action steps to reach its answers. For most subscribers using the Gemini app, this is invisible — you're not billed per token. For API developers, it's a real budget consideration.
What is Gemini 3.8 Flash Cyber, and can I use it?+
Gemini 3.8 Flash Cyber is a specialized variant built for cybersecurity tasks like vulnerability discovery and automated patching. It is not on general release — access is restricted to trusted defenders through Google's Fairwind Program. Standard Gemini users and developers access the regular 3.8 Flash model, not the Cyber variant.
Should I switch from 3.7 Flash to 3.8 Flash?+
Yes, for coding and agentic tasks — 3.8 Flash is meaningfully better. For general writing, summarization, or Q&A, the improvement is smaller, and 3.7 Flash may remain available at lower cost in some integrations. If you're a Gemini Pro or Ultra subscriber using the app, the switch is free and automatic where Google has rolled it out. API developers should test 3.8 Flash on their workloads and benchmark actual task costs before committing.
Related Guides
What Is Claude Fable 5.1, and Is It Actually Cheaper? (September 2026)
Anthropic released Claude Fable 5.1 on September 1, 2026, cutting cache-read token costs 75% and delivering up to 45% savings on agentic workflows. Here's what changed from Fable 5, who gets access, and whether you need to do anything today.
Can AI Improve Itself Now? What Anthropic's Automated Alignment Experiment Means for You
Anthropic published research on August 28, 2026 showing that nine Claude Opus 4.6 agents outperformed human alignment researchers on a specific AI safety task — the clearest proof yet that AI can meaningfully accelerate its own development. Here's what professionals using Claude need to understand about what happened and what didn't.
Are AI-Powered Cyberattacks on My Business a Real Risk? What 100 Companies Just Said
On August 27, 2026, OpenAI, Anthropic, Google, and 100+ other companies signed an open letter warning that AI-enabled cyberattacks will become 'far more widespread and sophisticated in a matter of months.' Here's what that means for professionals and three steps to take now.