Gemini 3.8 Flash Tested on a Real Board Deck

“Flash” suggests fast and cheap. On a multi-step analysis job it was neither the fastest nor the cheapest, and its deck covers the figures more than the stories behind them.

By Ed · Published November 5, 2026 · 5 min read

Twenty-five years starting companies, consulting, owning a multi-million-dollar P&L and leading business intelligence for large enterprises.

Gemini 3.8 Flash built a 10-slide board review for $1.47 in 13.8 minutes, the slowest of the seven models that finished. It covers the headline figures broadly: wholesale wins, Google campaigns, Meta audiences and placements. It did not follow the two biggest stories in the data, and found 1 of 8 story points in our checklist.

How this was tested. One prompt, “Northvale 2026 year-to-date board review, with an executive summary and recommendations”, run once per model in the Quiriz PowerPoint add-in on November 5, 2026. Northvale is a fictional US apparel company with stories planted in its data: Google and Meta ad exports and a HubSpot wholesale pipeline, through September 15, 2026. Every step ran on the model under test: plan, build (answering each slide’s question from the data), gap analysis, write and critic; see how a deck is built. Three small parts stayed fixed for every run (chart styling, question routing and a grounding check). Cost is what OpenRouter billed. Build time runs from pressing Build to the last slide written, and leaves out the plan, which waits for a person to approve it. One run per model, so treat close results as a tie. Full method on the benchmark page.

The run

Billed cost	$1.47 ($1.18 answers, $0.12 writing and critique, $0.06 plan)
Build time	13.8 minutes (slowest of seven)
Analysis calls	59, reading 2.7 million input tokens
Slides	10 content slides
Stories found	1 of 8 (see the checklist)

What it covered

A KPI row of Google Ads spend, Meta spend, wholesale won amount and deals won.
Wholesale won amount up 94.1% on last year, month by month.
Brand-Search as the biggest source of Google conversion value, about 42% of the total.
Retargeting-DPA returning $17,989 in conversion value on $4,221 of Meta spend, and Stories as the strongest placement.
The Southeast as the region lagging the rest.

Gemini 3.8 Flash deck: wholesale deal revenue grew 94.1% over last year — Wholesale won amount by month, this year against last, from the Gemini 3.8 Flash deck.

Gemini 3.8 Flash deck: Facebook ad cost and return by audience group — Meta cost and conversion value by audience, from the Gemini 3.8 Flash deck.

What it left out

The Meta prospecting decline. From May, Prospecting-Lookalike returned less than $2 per $1 spent, and in June less than $1. The deck covers its spend but not that decline.
Why the Southeast lags. Its win rate fell to 20% while every other region held above 50%, mostly on price. The deck shows the region behind, but not the reason.

Its recommendations

Four: sales coverage on the Midwest and the West, search budget on Brand-Search, and more Meta retargeting. All four back what is already working. None addresses the prospecting campaign or the Southeast, which is where the money is.

Why “Flash” was slow

Each answer is a small loop of queries. Flash took more rounds per question than any other model, and wrote 50,000 tokens in its answers against Sol’s 8,000. Fast calls, many of them. If speed matters, measure the task, not the token rate.

Verdict

Good for a quick tour of the numbers, light for a board deck. For about twice the price, Gemini 4 Argon found 6 of 8 story points on the same job; for a fifth of the price, DeepSeek V4.1 Flash found the ad problem.

Build a board deck from your own data

Describe the deck in PowerPoint and see the source of any figure. Free to start.

Try Quiriz free →

Frequently asked questions

Is Gemini 3.8 Flash good for data analysis?

For a summary of totals and campaign returns, yes. In this test it went less deep than the larger models: it did not follow the Meta prospecting decline or the regional win rates, which are the two biggest stories in the data.

How much did the Gemini 3.8 Flash deck cost?

$1.47 billed through OpenRouter for the whole run, of which $1.18 was answering the data questions. Its list price is $0.83 per million input tokens and $4.12 per million output tokens.

Is Gemini 3.8 Flash fast?

Per call, yes. Per task, not here: it took 13.8 minutes, the slowest of seven models, because it made 59 analysis calls where GPT-6.1 Sol made 33.

How does Gemini 3.8 Flash compare with Gemini 4 Argon?

Argon cost about twice as much ($2.98) and took about as long, but found 6 of the 8 story points in our checklist against Flash’s 1, with deeper drill-downs and specific recommendations.