GPT-5.5 vs Kimi K3 for multi-step research
GPT-5.5 wins, 91–88
Same 12 multi-step research tasks, marked blind by three rival labs. GPT-5.5 took 6 tasks, Kimi K3 took 1, 5 tied. Tested 18 Aug 2026.
Where they differed most
A charity trustee asks: 'should our shop open on Sundays?' That is a decision, not a research question. Turn it into exactly 5 sub-questions, numbered 1-5, one line each and no sub-points. Each line must name (a) the sub-question, (b) the kind of source or record that would answer it — the shop's own till data, the rota and wage cost, the lease or planning conditions, comparable local shops, the volunteer pool — and (c) the finding that would flip the decision. No general research advice: five specific, answerable questions.
1. Sunday revenue potential — the shop’s own till data — if nearby bank-holiday or extended-hours sales suggest Sunday takings would exceed normal weekday takings, flip toward opening. 2. Net staffing cost — the rota and wage cost — if Sunday cover would require paid hours that wipe out expected surplus, flip toward staying closed. 3. Permission to trade — the lease or planning conditions — if Sunday trading is prohibited or requires costly consent, flip toward staying closed. 4. Local demand benchmark — comparable local shops — if simila…
1. Is there proven weekend demand worth opening for? — read the shop's own till data by day of week; the decision flips to "no" if Saturday, the closest proxy, does not already beat the weekday average. 2. Would Sunday pay for its own staffing? — cost the extra Sunday shifts from the rota and wage cost records against projected Sunday takings; the decision flips to "no" if wages would consume the entire margin. 3. Are we actually permitted to trade on Sundays? — check the lease or planning conditions; the decision flips to "no" outright if eith…
Task by task
| Task | GPT-5.5 | Kimi K3 |
|---|---|---|
| Four numbers, one high street | 9 | 9 |
| The trial that both worked and didn't | 9 | 9 |
| Break the brief into answerable steps | 8 | 9 |
| Corroboration that isn't | 9 | 9 |
| The disagreement is definitional | 9 | 8 |
| A forecast is not an outcome | 10 | 9 |
| The evidence table, as JSON | 9 | 8 |
| Funded, and still the best data | 9.5 | 9 |
| Write only the limits section | 9 | 8 |
| Not answerable from what you have | 10 | 9 |
| The statistic and the doorstep | 9 | 9 |
| Refuse the manufactured briefing | 9 | 9 |
Full receipts: GPT-5.5, Kimi K3 · judges claude-sonnet-5, gemini-3.1-pro-preview, grok-4.5
Questions people ask
Which is better for multi-step research: GPT-5.5 or Kimi K3?
GPT-5.5 — it scored 91/100 against 88/100 on our 12-task multi-step research suite, winning 6 tasks to 1 with 5 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-5.5 · Claude Sonnet 5 vs Kimi K3 · GPT-5.5 vs GPT-5.6 Sol · GPT-5.5 vs Grok 4.5 · Claude Opus 4.8 vs GPT-5.5 · Gemini 3.1 Pro Preview vs GPT-5.5
Full ranking: Best AI for multi-step research · model pages: GPT-5.5, Kimi K3