Best AI for coding / head-to-head

GPT-5.3-Codex vs GPT-5.6 Terrafor coding

The verdict

GPT-5.6 Terra wins, 9392

Same 18 coding tasks, marked blind by three rival labs. GPT-5.6 Terra took 3 tasks, GPT-5.3-Codex took 2, 13 tied. Tested 16 Aug 2026.

Where they differed most

The task — Duration parser

Write a Python function parse_duration(s) that converts strings like '1h30m', '45s', '2h', '90m', '1h2m3s' into total seconds. Raise ValueError on anything malformed. Include three assert-based tests, one of which covers a malformed input.

```python
import re

def parse_duration(s: str) -> int:
    """
    Convert duration strings like:
      '1h30m', '45s', '2h', '90m', '1h2m3s'
    into total seconds.

    Rules:
    - Units must be in h -> m -> s order if present.
    - Each unit can appear at most once.
    - At least one unit must be present.
    - No extra characters/spaces allowed.
    """
    if not isinstance(s, str):
        raise ValueError("Duration must be a string")

    m = re.fullmatch(r'(?:(\d+)h)?(?:(\d+)m)?(?:(\d+)s)?', s)
    if not m:
        raise ValueError…
```python
import re

_DURATION_RE = re.compile(r"(?:(\d+)h)?(?:(\d+)m)?(?:(\d+)s)?")

def parse_duration(s):
    """Convert a duration like '1h30m' into total seconds."""
    if not isinstance(s, str):
        raise ValueError("Duration must be a string")

    match = _DURATION_RE.fullmatch(s)
    if not match or not any(match.groups()):
        raise ValueError(f"Malformed duration: {s!r}")

    hours, minutes, seconds = (
        int(value) if value is not None else 0
        for value in match.groups()
    )

    return hours * 3600 + minute…

Task by task

TaskGPT-5.3-CodexGPT-5.6 Terra
Duration parser98
Find the bug1010
SQL without window functions1010
Infinite useEffect910
Typed debounce1010
Leftmost binary search1010
Security review910
Safe migration89
Fix the code not the test109
Race condition99
Retry with backoff88
Recursive type99
Bash one-liner99
Explain and cost1010
Regex with limits99
Callback to async99
Diagnose from a trace88
Idempotency1010

Full receipts: GPT-5.3-Codex, GPT-5.6 Terra · 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-5.6 Terra?

GPT-5.6 Terra — it scored 93/100 against 92/100 on our 18-task coding suite, winning 3 tasks to 2 with 13 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-5.6 Luna vs GPT-5.6 Terra · GPT-5.3-Codex vs GPT-5.6 Luna · GPT-5.6 Sol vs GPT-5.6 Terra · GPT-5.3-Codex vs GPT-5.6 Sol · GLM 5.2 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-5.6 Terra