Head to head
vs Claude Sonnet 4 is retired.
For light moderate heavy
No verdict
One cost is modelled and the other is stale . The two bases are never ranked against each other, so no verdict is given. Each figure is shown on its own basis below.
Copy link Save scenario Export compare table Planned
Verified pricing, pass rate and cost per solved task for Gemini 3.1 Pro (preview) and Claude Sonnet 4 Figure Gemini 3.1 Pro (preview) Claude Sonnet 4 Input $/Mtok $2 $3 Output $/Mtok $12 $15 Cached input $/Mtok $0.2 $0.3 Pass rate 46% 61% Cost per solved task $1.91 modelled $1.96 stale Cost per month $382 $392
Cost per month against tasks per month · drag the marker
Monthly cost against tasks per month Log-log lines for Gemini 3.1 Pro (preview) (modelled, $1.91 per solved task), Claude Sonnet 4 (stale, $1.96 per solved task) from 10 to 100,000 tasks a month. Marker at 200 tasks. 10 100 1k 10k 100k $100 $1k $10k $100k TASKS / MONTH (log) → 200 tasks Gemini 3.1 Pro (preview) · MODELED · $1.91 per solved task $382 Gemini 3.1 Pro (prev… MODELED Claude Sonnet 4 · STALE · $1.96 per solved task $392 Claude Sonnet 4 STALE Monthly cost against tasks per month Log-log lines for Gemini 3.1 Pro (preview) (modelled, $1.91 per solved task), Claude Sonnet 4 (stale, $1.96 per solved task) from 10 to 100,000 tasks a month. Marker at 200 tasks. 10 100 1k 10k 100k $100 $1k $10k $100k TASKS / MONTH (log) → 200 tasks Gemini 3.1 Pro (preview) · MODELED · $1.91 per solved task $382 Claude Sonnet 4 · STALE · $1.96 per solved task $392 — Gemini 3.1 Pro (preview) · MODELED — Claude Sonnet 4 · STALE Solid measured · dashed modeled · 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