Best AI for coding / head-to-head
GPT-5.3-Codex vs GPT-5.5for coding
GPT-5.3-Codex wins, 92–91
Same 18 coding tasks, marked blind by three rival labs. GPT-5.3-Codex took 6 tasks, GPT-5.5 took 5, 7 tied. Tested 16 Aug 2026.
Where they differed most
Write an async retry wrapper in TypeScript: exponential backoff with jitter, a maximum attempt count, and it must NOT retry on 4xx responses other than 429. Maximum 30 lines. State what happens on the final failure.
```ts
type RetryOptions = {
maxAttempts: number;
baseDelayMs?: number;
maxDelayMs?: number;
jitterRatio?: number; // 0.2 => ±20%
};
export async function retryWithBackoff<T>(
op: () => Promise<T>,
{ maxAttempts, baseDelayMs = 200, maxDelayMs = 10_000, jitterRatio = 0.2 }: RetryOptions
): Promise<T> {
for (let attempt = 1; attempt <= maxAttempts; attempt++) {
try {
return await op();
} catch (err: any) {
const status = err?.status ?? err?.response?.status;
const nonRetriable4xx = status >= 400 && status <…```ts
type RetryOptions = { attempts?: number; baseMs?: number; maxMs?: number };
const sleep = (ms: number) => new Promise(r => setTimeout(r, ms));
const statusOf = (x: any): number | undefined =>
typeof x?.status === "number" ? x.status :
typeof x?.response?.status === "number" ? x.response.status :
undefined;
const isNonRetryable4xx = (s?: number) =>
s !== 429 && s !== undefined && s >= 400 && s < 500;
export async function retry<T>(
fn: () => Promise<T>,
{ attempts = 5, baseMs = 200, maxMs = 5000 }: RetryOptions = {}
): Prom…Task by task
| Task | GPT-5.3-Codex | GPT-5.5 |
|---|---|---|
| Duration parser | 9 | 8 |
| Find the bug | 10 | 10 |
| SQL without window functions | 10 | 10 |
| Infinite useEffect | 9 | 10 |
| Typed debounce | 10 | 9 |
| Leftmost binary search | 10 | 10 |
| Security review | 9 | 10 |
| Safe migration | 8 | 9 |
| Fix the code not the test | 10 | 9 |
| Race condition | 9 | 10 |
| Retry with backoff | 8 | 5 |
| Recursive type | 9 | 10 |
| Bash one-liner | 9 | 9 |
| Explain and cost | 10 | 10 |
| Regex with limits | 9 | 8 |
| Callback to async | 9 | 9 |
| Diagnose from a trace | 8 | 8 |
| Idempotency | 10 | 9 |
Full receipts: GPT-5.3-Codex, GPT-5.5 · judges claude-sonnet-5, gemini-3.1-pro-preview, grok-4.5
Questions people ask
Which is better for coding: GPT-5.3-Codex or GPT-5.5?
GPT-5.3-Codex — it scored 92/100 against 91/100 on our 18-task coding suite, winning 6 tasks to 5 with 7 tied. Every answer was marked blind by three judges from three rival AI labs.
How was this tested?
Both models answered the identical published coding 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 coding head-to-heads: GPT-5.3-Codex vs GPT-5.6 Luna · GPT-5.5 vs GPT-5.6 Luna · GPT-5.3-Codex vs GPT-5.6 Sol · GPT-5.5 vs GPT-5.6 Sol · GPT-5.3-Codex vs GPT-5.6 Terra · GPT-5.5 vs GPT-5.6 Terra
Full ranking: Best AI for coding · model pages: GPT-5.3-Codex, GPT-5.5