GPT-6.1 Sol Tested on a Real Board Deck

Sol reads like a careful analyst: every figure labeled, every limit stated. That makes it easy to trust and slow to commit, and the biggest sales story did not make its deck.

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.

GPT-6.1 Sol built an 11-slide review in 5.4 minutes for $2.08, the fastest of seven models. Every figure is carefully labeled and every comparison uses the same dates each year. It caught the rising cost of Meta prospecting, but did not look at win rates by region, so the Southeast story is missing. It found 3 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	$2.08 ($1.66 answers, $0.14 slide notes, $0.12 writing and critique)
Build time	5.4 minutes (fastest of seven)
Analysis calls	33 (fewest of seven), reading 1.4 million input tokens
Slides	11 content slides
Stories found	3 of 8 (see the checklist)

What it found

The Lookalike cost problem. Prospecting-Lookalike took 82.3% of the added Meta spend, cost up 50% while conversions stayed flat, and cost per conversion rose in every placement of both ad sets.
Google efficiency. NonBrand-Search conversions grew 21.4% on 12.1% more cost, while PMax-Retarget’s 165% cost rise outran its conversions.
Loss reasons over time. Pricing losses up $104.7K to $271.1K, and deals going dark up $150.8K, the largest increase.
Careful labels. Ad conversions are called platform-reported throughout, and the deck says what the selected data cannot answer.

What it left out

The Southeast. Sol ranked lost deal amount by region, found the regions close, and concluded that losses were too spread out to concentrate sales coverage. It never set losses against wins: the Southeast won 4 of its 20 closed deals while every other region won more than half. That is the biggest sales story in the data.

Its recommendations

Three, ranked: require value-and-price proposals on price-sensitive deals; require dated buyer commitments to stop deals going dark; run a controlled test moving Lookalike budget into NonBrand-Search. All three are reasonable and properly hedged. None targets the Southeast, and the media one stops short of cutting a campaign its own slides show losing efficiency.

GPT-6.1 Sol deck: ranked priorities: pricing, buyer commitments, then budget reallocation — GPT-6.1 Sol’s ranked recommendations, each with an owner.

The slide notes

The most precise of any model: each observation states its basis and its limit. The trade-off is tone. Phrases like “does not establish profitable incremental growth” are careful and leave a board without a verdict.

Verdict

The analyst you trust not to overstate, and the fastest. If you want it to find the problem and say what to do, Claude Opus 5.5 did both. For cheaper comparisons see GPT-6 Luna vs Sol and DeepSeek vs GPT-6.1 Sol.

Build a board deck from your own data

Describe the deck in PowerPoint and Quiriz answers each slide from your connected data. Free to start.

Try Quiriz free →

Frequently asked questions

Is GPT-6.1 Sol good for data analysis?

It is careful and fast: in this test it labeled every ad figure as platform-reported, compared the same dates each year, and said what the data could not answer. It was cautious about conclusions, and did not look at win rates by region, where the biggest sales story was.

How much did the GPT-6.1 Sol deck cost?

$2.08 billed through OpenRouter for the whole run, $1.66 of it answering the data questions. List price is $2 per million input tokens and $10 per million output tokens; caching brought the answers in 41% under list.

How fast is GPT-6.1 Sol?

5.4 minutes from pressing Build to the last slide, the fastest of seven models. It made 33 analysis calls, the fewest of any model.

Is GPT-6.1 Sol the model in ChatGPT?

We tested the GPT-6.1 Sol model through OpenRouter’s API, not the ChatGPT app. Which model ChatGPT uses for a given conversation is up to OpenAI.