Gemini 3.5 Flash vs GPT-5.6 Sol for multi-step research

The verdict

GPT-5.6 Sol wins, 8978

Same 12 multi-step research tasks, marked blind by three rival labs. GPT-5.6 Sol took 10 tasks, Gemini 3.5 Flash took 1, 1 tied. Tested 18 Aug 2026.

Where they differed most

The task — The disagreement is definitional

Three sources on the average deposit a UK first-time buyer needs. Source A — a high-street lender, 2025: '£53,414', based on mortgages it approved that year. Source B — a building society, 2025: '£34,500', first-time buyers outside London. Source C — a London estate agency, 2024: '£71,000'. These are usually written up as a contradiction. Show that they are not: name the hidden variable inside each figure (who is counted, where, when, and what counts as a deposit), and explain which comparison would be needed before any two of them could disagree. Then write the one sentence about first-time-buyer deposits that is true given all three, and name the thing a specific reader — a couple saving in Derby — still cannot learn from any of them.

These three figures are often presented as a contradiction, but they are actually measuring entirely different things. Here is the breakdown of the hidden variables that explain the variation, followed by what is needed for a true comparison, the single true sentence, and the missing information for the Derby couple.

### The Hidden Variables

To understand why these figures coexist peacefully, we must look at the four hidden variables behind each number:

#### Source A (£53,414)
*   **Who is counted:** Only first-time buyers who successfully a…
“Average deposit” is not a single, universal quantity. Each figure is a conditional average:

| Source | Who is counted | Where | When | What “deposit” means |
|---|---|---|---|---|
| **A: £53,414** | First-time buyers whose applications were approved by that particular lender—not rejected applicants, cash buyers or customers of other lenders | The lender’s own geographic mix, which need not represent the UK evenly | Approvals recorded in 2025, rather than necessarily purchases completed in 2025 | Usually the proposed purchase price minus the a…

Task by task

TaskGemini 3.5 FlashGPT-5.6 Sol
Four numbers, one high street810
The trial that both worked and didn't89
Break the brief into answerable steps810
Corroboration that isn't89
The disagreement is definitional69
A forecast is not an outcome89
The evidence table, as JSON78
Funded, and still the best data79
Write only the limits section99
Not answerable from what you have910
The statistic and the doorstep85
Refuse the manufactured briefing810

Full receipts: Gemini 3.5 Flash, GPT-5.6 Sol · judges claude-sonnet-5, gpt-5.6-terra, grok-4.5

Questions people ask

Which is better for multi-step research: Gemini 3.5 Flash or GPT-5.6 Sol?

GPT-5.6 Sol — it scored 89/100 against 78/100 on our 12-task multi-step research suite, winning 10 tasks to 1 with 1 tied. Every answer was marked blind by three judges from three rival AI labs.

How was this tested?

Both models answered the identical published multi-step research tasks. Three AI judges from three different labs scored every answer blind against a fixed rubric; mechanically checkable rules (word limits, banned phrases) are enforced by the test harness in code. The raw outputs and judge verdicts are downloadable.

More multi-step research head-to-heads: Claude Sonnet 5 vs GPT-5.6 Sol · Claude Sonnet 5 vs Gemini 3.5 Flash · GPT-5.5 vs GPT-5.6 Sol · Gemini 3.5 Flash vs GPT-5.5 · GPT-5.6 Sol vs Kimi K3 · GPT-5.6 Sol vs Grok 4.5

Full ranking: Best AI for multi-step research · model pages: Gemini 3.5 Flash, GPT-5.6 Sol