Gemini 3.8 Flash Launches at $0.75/1M Tokens [2026]

Google shipped Gemini 3.8 Flash on September 2, 2026, pricing the model at $0.75 per million input tokens and $3.75 per million output tokens, the same introductory rate the company used for Gemini 3.7 Flash just three weeks earlier. The launch, confirmed on Google’s official blog and first reported by 9to5Google, marks the fourth Flash-tier release in four months and extends a pricing standoff that is now reshaping how developers budget for AI inference. Google calls it “our most intelligent Flash model,” built for long-horizon software engineering, autonomous agents, and multi-step enterprise reasoning.

The timing matters. Tech Insider covered the earlier speculation that Gemini 3.8 Flash could land within days and the subsequent leaked benchmark numbers that hinted at an 80 percent price war across the AI sector. Both stories are now resolved: the model is real, it is live, and the pricing undercuts most rivals by a wide margin. What follows is a full breakdown of the numbers, the competitive picture, and what happens once the introductory pricing window closes on December 31, 2026.

Google Ships Gemini 3.8 Flash Three Weeks After 3.7 Flash

Gemini 3.7 Flash launched in mid-August 2026. Gemini 3.8 Flash followed on September 2, a gap of roughly three weeks, according to 9to5Google’s launch coverage. That cadence is unusually fast even by 2026 standards, where major labs have been shipping incremental model updates every few weeks rather than the multi-month cycles that defined the 2023-2024 era. Google’s own documentation lists Gemini 3.8 Flash, Gemini 3.7 Flash, and Gemini 3.6 Flash as all carrying the same introductory pricing window through the end of the year, which suggests Google is treating the Flash line less like discrete product launches and more like a rolling release train.

The model is available immediately through the generateContent API, Google AI Studio, and the Gemini Enterprise Agent Platform. It has also rolled into consumer-facing surfaces: the Gemini app for Google AI Pro and Ultra subscribers, AI Mode in Search, and Gemini inside Google Sheets. That breadth of distribution on day one is notable. Earlier Flash releases sometimes took days or weeks to reach every surface. This one launched everywhere at once.

What $0.75 per Million Input Tokens Actually Buys

At $0.75 per million input tokens and $3.75 per million output tokens, a developer processing 10 million input tokens and generating 2 million output tokens in a month would pay roughly $15 total, a workload that would have cost several times more on a flagship-tier model just a year ago. The pricing is not a discount tier or a limited-time trial gimmick in the traditional sense. It is the standard rate through December 31, 2026, after which it steps up to $1.50 per million input tokens and $7.50 per million output tokens starting January 1, 2027, per Google’s official pricing documentation.

On the Gemini Enterprise Agent Platform, the numbers get more interesting for high-volume users. Cached input tokens are priced at $0.075 per million through the end of 2026, a tenth of the standard input rate, according to Google Cloud’s pricing page. That cached rate rises to $0.15 per million starting in January 2027. For agentic workloads that repeatedly reference the same system prompts, tool definitions, or long documents, caching turns Gemini 3.8 Flash into one of the cheapest inference options on the market right now.

Two Models, One Launch: Gemini 3.8 Flash and 3.8 Flash Cyber

Google’s announcement bundled two models under one release: Gemini 3.8 Flash and a variant called Gemini 3.8 Flash Cyber, per the title of Google’s own blog post. The general-purpose Flash model is positioned for coding, agent workflows, and enterprise reasoning tasks. Google has not published a detailed technical breakdown separating the two models’ capabilities beyond the launch post itself, so specifics on what differentiates the Cyber variant remain limited at this stage.

Why Bundle Two Releases Together

Pairing a general model with a specialized variant in a single announcement is a pattern Google has leaned on before, and it lets the company address both broad developer demand and narrower enterprise use cases without staggering separate launch cycles. It also means both models inherit the same introductory pricing window, simplifying the rollout for Google’s own product teams as much as for outside developers.

Gemini 3.8 Flash Pricing vs the Rest of the Gemini Lineup

Placed next to its own siblings, Gemini 3.8 Flash undercuts Google’s higher-end Pro tier by a wide margin while matching the immediate prior Flash release on headline price. The table below lays out the current lineup based on Google’s published pricing documentation.

Model	Input ($/1M tokens)	Output ($/1M tokens)	Pricing valid through	Next-tier rate (from Jan 1, 2027)
Gemini 3.8 Flash	$0.75	$3.75	Dec 31, 2026	$1.50 / $7.50
Gemini 3.7 Flash	$0.75	$3.75	Dec 31, 2026	$1.50 / $7.50
Gemini 3.6 Flash	$1.50	$7.50	Standard rate	N/A
Gemini 3.1 Pro	~$2.00	~$12.00	Standard rate	N/A

The pattern is straightforward: each new Flash release launches at the discounted rate the previous one is about to lose, effectively keeping one generation of Flash permanently at the low introductory price while the model behind it ages into standard pricing. Readers who want the fuller picture on how Gemini 3.7 Flash’s own price cut played out can see the same mechanic in action one cycle earlier.

The Pareto Frontier Claim: What Artificial Analysis Found

Independent benchmarking firm Artificial Analysis scored Gemini 3.8 Flash at 59 on its Intelligence Index and placed the model on what it calls the “Pareto frontier” for intelligence versus cost per task, meaning no other model available at a similar or lower price currently scores higher on the firm’s composite reasoning benchmark. That assessment, visible on Artificial Analysis’s own model page, is one of the few third-party data points available so early in the launch cycle.

Google has not yet published a full benchmark suite specific to Gemini 3.8 Flash covering coding, math, and multimodal reasoning separately. The closest official reference points come from the preceding Gemini 3 Flash preview, which posted a 100.0 percent score on Vending-Bench 2, 99.7 percent on AIME 2025, 92.8 percent on Global PIQA, 91.8 percent on MMMLU, and 90.4 percent on GPQA, according to benchmark tracker llm-stats. Those figures describe the prior model generation, not 3.8 Flash directly, and should be read as an indication of the Flash family’s trajectory rather than a confirmed score for the new release.

In Google’s Own Words

“It is available at the same introductory price as 3.7 Flash at $0.75 per million input tokens and $3.75 per million output tokens.”

Google, official launch blog

That line, published on Google’s launch post, is the closest thing to an official statement on why the company chose price parity with the prior release rather than a further cut or a premium markup for the newer, more capable model. Google’s developer documentation reinforces the same point: developers can “take advantage of introductory pricing across Google AI Studio and Gemini Enterprise Agent Platform through December 31, 2026 for Gemini 3.8 Flash, Gemini 3.7 Flash, and Gemini 3.6 Flash,” according to the official Gemini API docs.

9to5Google’s launch-day coverage framed the move in blunt terms, noting that “Google is once again offering an introductory price of $0.75/1M input tokens and $3.75/1M output tokens until December 31,” in its report on the release. The repetition of the word “again” is doing some work there: this is now the third consecutive Flash launch to open at the exact same headline rate.

How Gemini 3.8 Flash Stacks Up Against GPT-5.6, Claude and Qwen

Zoom out to the broader market and Gemini 3.8 Flash looks aggressively priced against nearly every major rival except the cheapest open-weight options. OpenAI’s GPT-5.6 Sol sits at $4.00 input and $20.00 output per million tokens on OpenAI’s own pricing page, more than five times Gemini 3.8 Flash’s input rate. The mid-tier GPT-5.6 Terra comes in at $2.00 and $12.00. OpenAI’s budget option, GPT-5.6 Luna, undercuts Gemini at $0.20 input and $1.20 output, confirmed by OpenAI’s pricing page and its own announcement.

Anthropic’s flagship pricing sits well above Google’s Flash tier. Claude Opus 5 is confirmed at $5.00 input and $25.00 output per million tokens, while Claude Fable 5 carries a $10.00 input and $50.00 output rate, both figures confirmed directly by Anthropic. On the cheaper end, DeepSeek’s V4-Flash publishes a base cache-miss rate of $0.14 input and $0.28 output per million tokens on its own pricing page, undercutting Gemini 3.8 Flash outright. Alibaba’s Qwen3.8-Flash-Next is reported by Reuters at roughly 1 yuan input and 3 yuan output per million tokens, which converts to figures in the same low range as DeepSeek, though that USD conversion is derived from the yuan price rather than a separately published Alibaba figure.

Model	Vendor	Input ($/1M tokens)	Output ($/1M tokens)
Gemini 3.8 Flash	Google	$0.75	$3.75
GPT-5.6 Sol	OpenAI	$4.00	$20.00
GPT-5.6 Terra	OpenAI	$2.00	$12.00
GPT-5.6 Luna	OpenAI	$0.20	$1.20
Claude Opus 5	Anthropic	$5.00	$25.00
Claude Fable 5	Anthropic	$10.00	$50.00
DeepSeek V4-Flash	DeepSeek	$0.14	$0.28
Qwen3.8-Flash-Next	Alibaba	~$0.15 (derived from yuan price)	~$0.45 (derived from yuan price)

The picture that emerges is a market splitting into three tiers: expensive flagship reasoning models from OpenAI and Anthropic priced for maximum capability, a mid-tier occupied by Gemini’s Flash line that trades some ceiling for a sharply lower bill, and a cut-rate bracket led by DeepSeek and Qwen that undercuts everyone on raw token price. For a deeper look at how that bottom tier compares on pricing specifically, Tech Insider’s earlier breakdown of DeepSeek V4-Flash, Gemini 3.7 Flash, and Qwen3.8-Flash-Next covers the mechanics in more detail, and much of that comparison still holds with 3.8 Flash’s matching price point.

The Flash Release Cadence: Four Models in Four Months

Gemini 3.8 Flash is the fourth Flash-branded model Google has shipped in roughly four months, following Gemini 3.6 Flash, Gemini 3.7 Flash, and now this release. A market report cited by 24/7 Wall St framed the push as occurring “at the AI cost frontier,” language that captures what Google appears to be optimizing for: not necessarily the single smartest model on the market, but the smartest model available at each specific price point, updated on a near-monthly basis.

That cadence puts pressure on competitors in a way that a single annual flagship release cannot. Every few weeks, Google resets the bar for what “cheap and capable” means, which forces rivals to either match the price cut or defend a widening capability gap. It also puts pressure on Google’s own infrastructure teams, since each release has to clear safety review, capacity planning, and rollout across consumer and enterprise surfaces on a compressed timeline.

Why Google Keeps Undercutting Its Own Prior-Gen Pricing

Google Cloud’s own pricing page confirms that Gemini 3.8 Flash, 3.7 Flash, and 3.6 Flash have all shared the same introductory pricing structure at various points, a pattern that looks less like a series of one-off promotions and more like a deliberate floor Google has set for its mid-tier model line. The logic is not complicated: developer mindshare in the AI infrastructure market is sticky. Whoever a team builds its agent pipeline, coding assistant, or customer-support bot around today is unlikely to migrate providers casually once the integration work is done. Locking in that mindshare at a price rivals struggle to match is a land-grab strategy, not a charity discount.

There is also a defensive angle. DeepSeek and Qwen’s ultra-low pricing has already pulled meaningful developer attention toward open-weight and China-based providers over the past year, according to the broader pricing landscape reflected in industry trackers. Google matching or beating that price band with a proprietary, fully-hosted, enterprise-supported model gives budget-conscious developers a reason to stay inside Google’s ecosystem rather than route workloads elsewhere.

What’s Still Unconfirmed About Gemini 3.8 Flash

Not every detail of the launch is nailed down. As of September 2, 2026, Google had not published a dedicated context-window specification or a full capability list specific to Gemini 3.8 Flash beyond the launch blog itself, a gap noted by industry commentary the same day. The immediately preceding Gemini 3 Flash preview carried a 1 million token input context window and a 65,536 token output context window, and the Flash family generally has been built around long-context design, but that figure has not been separately confirmed for the 3.8 release specifically.

Context Window Ambiguity

This is a meaningful gap for developers deciding whether to migrate a production workload today versus waiting for fuller documentation. A model can be cheap per token and still be expensive in practice if its usable context is smaller than advertised or than a competing model, since that forces more chunking, more retrieval overhead, and more round-trip calls. Until Google publishes 3.8-specific numbers, teams evaluating a migration should test against their own workloads rather than assume parity with the prior Flash generation.

Historical Context: From Gemini 1.5 Flash to Gemini 3.8

Google introduced the original Flash concept years ago as a lightweight, lower-cost counterpart to its flagship Pro models, aimed at high-volume tasks where raw reasoning depth mattered less than throughput and price. What has changed by 2026 is the pace and the ambition of each Flash release. Early Flash models were explicitly positioned as the budget option with real capability trade-offs. Gemini 3.8 Flash, by contrast, is marketed as Google’s most intelligent Flash model to date, aimed at software engineering and multi-step agent workflows that would have required a Pro-tier model just a generation or two ago. The Flash line has effectively absorbed capability that used to be reserved for premium tiers, while keeping the price closer to the original budget positioning.

That shift mirrors a broader trend across the industry that Tech Insider has tracked through 2026, including the expansion of free and low-cost tiers across ChatGPT, Claude, and Gemini. Capability that once justified a premium price tag is steadily migrating down-market as competition intensifies and inference costs fall industry-wide.

What Developers and Enterprises Should Do Before December 31

Teams currently building on Gemini 3.7 Flash or 3.6 Flash have a decision to make well before the introductory pricing window closes. Migrating to Gemini 3.8 Flash costs nothing extra at today’s rate, since both models share the same $0.75/$3.75 pricing, but it comes with the added benefit of what Google is calling stronger performance on coding and agentic tasks. For teams with high-cache workloads, specifically agent pipelines that repeatedly reuse system prompts or reference documents, testing the Enterprise Agent Platform’s $0.075 per million cached-token rate could meaningfully cut monthly bills before that rate also rises on January 1, 2027.

curl https://generativelanguage.googleapis.com/v1beta/models/gemini-3.8-flash:generateContent 
  -H "x-goog-api-key: $GEMINI_API_KEY" 
  -H "Content-Type: application/json" 
  -X POST 
  -d '{"contents":[{"parts":[{"text":"Summarize this quarter's inference costs."}]}]}'

Enterprises evaluating a multi-vendor strategy should also run head-to-head cost simulations against DeepSeek V4-Flash and Qwen3.8-Flash-Next before committing, since both remain cheaper on raw per-token price even if they carry different support, compliance, and latency profiles than a Google-hosted model.

Five Predictions for the Next Round of the AI Pricing War

Google will likely ship another Flash-tier update within weeks rather than months, continuing the three-week cadence established between 3.6, 3.7, and 3.8.
Expect at least one rival lab, most plausibly OpenAI or Anthropic, to introduce a new budget-tier model or price cut in direct response before the end of 2026, extending the pattern already visible with GPT-5.6 Luna’s aggressive pricing.
The December 31, 2026 pricing cliff will likely trigger a wave of migration and renegotiation activity as developers currently on 3.6, 3.7, or 3.8 Flash decide whether to lock in usage or wait for the next introductory-priced release.
Cached-token pricing, not headline input/output rates, will become the more important competitive battleground for enterprise agent platforms over the next two to three quarters.
Google will likely publish fuller context-window and benchmark documentation for Gemini 3.8 Flash within days to weeks of this launch, following the same pattern of delayed technical disclosure seen with 3.7 Flash.

Frequently Asked Questions

How much does Gemini 3.8 Flash cost per million tokens?

Gemini 3.8 Flash costs $0.75 per million input tokens and $3.75 per million output tokens at its introductory rate, valid through December 31, 2026, according to Google’s official pricing documentation.

When does Gemini 3.8 Flash’s introductory pricing expire?

The introductory pricing expires on December 31, 2026. Starting January 1, 2027, the rate rises to $1.50 per million input tokens and $7.50 per million output tokens.

Is Gemini 3.8 Flash cheaper than GPT-5.6 or Claude Opus 5?

Yes, on headline price. Gemini 3.8 Flash’s $0.75/$3.75 rate undercuts GPT-5.6 Sol ($4.00/$20.00), GPT-5.6 Terra ($2.00/$12.00), Claude Opus 5 ($5.00/$25.00), and Claude Fable 5 ($10.00/$50.00). It is more expensive than DeepSeek V4-Flash ($0.14/$0.28) and Qwen3.8-Flash-Next, both of which are priced lower.

What is Gemini 3.8 Flash Cyber?

Gemini 3.8 Flash Cyber is a companion model Google announced alongside Gemini 3.8 Flash in the same launch post. Google has not published a detailed breakdown separating its capabilities from the standard Flash model beyond the joint announcement.

Where is Gemini 3.8 Flash available?

It is available through the generateContent API, Google AI Studio, and the Gemini Enterprise Agent Platform, as well as in the Gemini app for Google AI Pro and Ultra subscribers, AI Mode in Search, and Gemini in Google Sheets.

What is Gemini 3.8 Flash’s context window?

Google has not published a context-window specification specific to Gemini 3.8 Flash as of this article’s publication. The prior Gemini 3 Flash preview model carried a 1 million token input context window, and the Flash family broadly is built around long-context design, but that figure is not yet formally confirmed for 3.8 Flash.

Did Gemini 3.8 Flash’s launch affect Alphabet’s stock price?

Market coverage tied to the launch noted a roughly 0.7 percent intraday rise in Alphabet (GOOGL) shares, framed as part of the broader four-models-in-four-months push at the low end of the AI pricing market. That figure reflects a single trading session and should not be read as a durable valuation shift tied solely to this release.