Skip to content
SOLVENCY

Head to head

GPT-5.4 vs Claude Opus 4.6

Claude Opus 4.6 is legacy.

For tasks, about a month.

Verdict · modelled basis

GPT-5.4 costs $1.86 per solved task against Claude Opus 4.6 at $3.85 — ▼ 2.1x cheaper. Over 200 tasks that is $398 a month.

On output token price alone GPT-5.4 looks 1.7x cheaper. The ranking holds, but the margin changes.

Verified pricing, pass rate and cost per solved task for GPT-5.4 and Claude Opus 4.6
FigureGPT-5.4Claude Opus 4.6
Input $/Mtok$2.5$5
Output $/Mtok$15$25
Cached input $/Mtok$0.25$0.5
Pass rate59%52%
Cost per solved task$1.86modelled$3.85modelled
Cost per month$372$771

Cost per month against tasks per month · drag the marker

Monthly cost against tasks per monthLog-log lines for GPT-5.4 (modelled, $1.86 per solved task), Claude Opus 4.6 (modelled, $3.85 per solved task) from 10 to 100,000 tasks a month. Marker at 200 tasks.101001k10k100k$100$1k$10k$100kTASKS / MONTH (log) →200 tasksGPT-5.4 · MODELLED · $1.86 per solved task$372Claude Opus 4.6 · MODELLED · $3.85 per solved task$771— GPT-5.4 · MODELLED— Claude Opus 4.6 · MODELLED

Solid measured · dashed modelled · dotted stale. The gap is the verdict.

Source: Scale SEAL leaderboard, SWE-bench Pro · verified 2026-08-21 · prices from each provider's pricing page; measured rows ignore the tier · Methodology · All models at these settings