GPT-5.3-Codex vs Kimi K3 for code review
GPT-5.3-Codex wins, 93–87
Same 12 code review tasks, marked blind by three rival labs. GPT-5.3-Codex took 5 tasks, Kimi K3 took 2, 5 tied. Tested 27 Aug 2026.
Where they differed most
Review this function. Context you can rely on, because it is validated upstream: `customers` and `outcodes` are always arrays; every customer object has a `postcode` property that is either a string or null; postcodes are stored either with or without the space ('SW1A 1AA' or 'SW1A1AA'); the required behaviour is to return the customers whose outward code matches one of `outcodes`, case-insensitively. function filterByOutcode(customers, outcodes) { const wanted = new Set(outcodes.map(o => o.trim().toUpperCase())); return customers.filter(c => { const pc = (c.postcode || '').trim().toUpperCase(); if (pc.length < 5) return false; const outward = pc.includes(' ') ? pc.split(' ')[0] : pc.slice(0, -3); return wanted.has(outward); }); } Give your review. If there are defects, list them by severity with the input that triggers each one. If there are none, say so plainly and briefly — do not manufacture a defect to look thorough, and do not dress a style preference up as a bug.
Given the constraints you provided, I don’t see a functional defect in this implementation. It correctly: - normalizes `outcodes` case-insensitively (`trim().toUpperCase()`), - normalizes customer postcodes the same way, - handles both stored formats (`"SW1A 1AA"` and `"SW1A1AA"`), - extracts outward code appropriately in each format, - returns only matching customers. So: **no defects to report under the validated upstream assumptions.**
**No defects.**
The function is correct under the stated guarantees. I checked the spots where bugs in this kind of code usually hide:
- **Null postcode**: `(c.postcode || '')` normalizes to `''`, which fails the length check and is excluded — correct, since a null postcode has no outward code.
- **No-space parsing**: `pc.slice(0, -3)` relies on the UK inward code always being exactly 3 characters, which holds. So `M11AA` → `M1`, `M101AA` → `M10`, `SW1A1AA` → `SW1A` — all correct for the 5/6/7-char forms.
- **Spaced parsing**: `pc.split(' ')[…Task by task
| Task | GPT-5.3-Codex | Kimi K3 |
|---|---|---|
| The pagination that skips page one | 8 | 9 |
| Nothing wrong with it | 10 | 5 |
| Three things wrong in six lines | 10 | 9 |
| Two customers, one last item | 9 | 9 |
| Rank by what breaks | 9 | 9 |
| The review, as JSON | 10 | 8 |
| Slow is not the same as wrong | 9 | 9 |
| Overdue at one in the morning | 8 | 9 |
| Write only the comment | 10 | 10 |
| The complaint that is not a defect | 9 | 9 |
| Right, and unusable | 10 | 9 |
| Refuse the rubber stamp | 10 | 9 |
Full receipts: GPT-5.3-Codex, Kimi K3 · judges claude-sonnet-5, gemini-3.1-pro-preview, grok-4.5
Questions people ask
Which is better for code review: GPT-5.3-Codex or Kimi K3?
GPT-5.3-Codex — it scored 93/100 against 87/100 on our 12-task code review suite, winning 5 tasks to 2 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 code review 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 code review head-to-heads: GPT-5.3-Codex vs GPT-5.6 Sol · GPT-5.6 Sol vs Kimi K3 · GPT-5.3-Codex vs GPT-5.6 Luna · GPT-5.6 Luna vs Kimi K3 · GPT-5.3-Codex vs GPT-5.6 Terra · GPT-5.6 Terra vs Kimi K3
Full ranking: Best AI for code review · model pages: GPT-5.3-Codex, Kimi K3