Claude Opus 4.8 vs Gemini 3.1 Pro Preview

Claude Opus 4.8 vs Gemini 3.1 Pro Preview

Independent runs now give Claude Opus 4.8 two real wins on agent work, each confirmed on both sides: Terminal-Bench 2.1 (78.9 vs 65.8) and Toolathlon-Verified (76.2 vs 61.1). Agents' Last Exam was a third until 2026-10-02 — its 27.0 vs 16.4 overall pass rate looked decisive, but Opus 4.8's row is the board's Max run while Gemini's is a High run, and more thinking is not a free upgrade, so that comparison no longer counts. The Terminal-Bench one is the weaker of the two: both rows are tbench.ai's, but with different agent scaffolds (Claude Code for Opus, Gemini CLI for Gemini), and vals.ai's single Terminus 2 harness puts them at 71.91 and 70.79, a tie. The reasoning and coding picture: HLE without tools is a statistical tie. GPQA Diamond and, since 2026-09-29, LiveCodeBench no longer belong in that group — both are graded saturated (grade D), so the close gaps there (92.42 vs 95.45, 87.82 vs 88.48) reflect ceiling-out tests, not evenly-matched models. (An earlier version of this page said Agents' Last Exam favored Gemini; that was our error — we had mixed the board's partial-credit score with its pass rate. Corrected 2026-08-17.) LiveBench moves the same way for the same reason (Max against High) and is no longer called either way, and AA-AnalystAgent, added 2026-09-29, another: 45.0 vs 41.25 pass^5, both run by Artificial Analysis, inside that benchmark's 11.2-point band. ARC-AGI-2 again doesn't yield a comparable line, since Opus 4.8's tracked score sits at its High tier (72.1) while Gemini's is a flat, untiered 77.1 — different conditions, no matched tier to compare. HMMT Feb 2026 doesn't add a usable line either: both scores (95.45 vs 94.70) carry MathArena's own disclosed contamination-warning, since every model this site tracks was released after the competition happened, so the whole benchmark is graded tainted rather than read as a real gap or a tie. Gemini's remaining case is price: $2/$12 vs $5/$25.

Claude Opus 4.8 vs Gemini 3.1 Pro Preview: benchmark by benchmark

Across 17 rows comparing Claude Opus 4.8 and Gemini 3.1 Pro Preview: 2 rows carry a confirmed real gap, 2 rows land inside the noise band, and 13 rows carry other caveats — unverified, tainted, or run on mismatched tool setups. Read each row’s Signal label before ranking Claude Opus 4.8 against Gemini 3.1 Pro Preview.

Who should pick Claude Opus 4.8, and who should pick Gemini 3.1 Pro Preview

  • Agentic work — terminal, multi-tool, professional tasks → Claude Opus 4.8 (two real gaps, every number on both sides independently run (tbench.ai, toolathlon.xyz); the Terminal-Bench one mixes agent scaffolds and ties on vals.ai's single harness, so lean on Toolathlon)
  • Budget-sensitive reasoning → Gemini 3.1 Pro Preview (ties Opus 4.8 on HLE no-tools (GPQA and LiveCodeBench are now excluded as saturated) at less than half the price)
  • Want every number verifiable → Claude Opus 4.8 (since independent boards covered both models, Opus 4.8's wins are the ones that survived — the earlier 'only Gemini is verified' framing is obsolete)

Claude Opus 4.8 vs Gemini 3.1 Pro Preview pricing

Input / 1M tokens$5.00$2.00
Output / 1M tokens$25.00$12.00

Claude Opus 4.8 vs Gemini 3.1 Pro Preview FAQ

Is Claude Opus 4.8 better than Gemini 3.1 Pro Preview?

Independent runs now give Claude Opus 4.8 two real wins on agent work, each confirmed on both sides: Terminal-Bench 2.1 (78.9 vs 65.8) and Toolathlon-Verified (76.2 vs 61.1). Agents' Last Exam was a third until 2026-10-02 — its 27.0 vs 16.4 overall pass rate looked decisive, but Opus 4.8's row is the board's Max run while Gemini's is a High run, and more thinking is not a free upgrade, so that comparison no longer counts. The Terminal-Bench one is the weaker of the two: both rows are tbench.ai's, but with different agent scaffolds (Claude Code for Opus, Gemini CLI for Gemini), and vals.ai's single Terminus 2 harness puts them at 71.91 and 70.79, a tie. The reasoning and coding picture: HLE without tools is a statistical tie. GPQA Diamond and, since 2026-09-29, LiveCodeBench no longer belong in that group — both are graded saturated (grade D), so the close gaps there (92.42 vs 95.45, 87.82 vs 88.48) reflect ceiling-out tests, not evenly-matched models. (An earlier version of this page said Agents' Last Exam favored Gemini; that was our error — we had mixed the board's partial-credit score with its pass rate. Corrected 2026-08-17.) LiveBench moves the same way for the same reason (Max against High) and is no longer called either way, and AA-AnalystAgent, added 2026-09-29, another: 45.0 vs 41.25 pass^5, both run by Artificial Analysis, inside that benchmark's 11.2-point band. ARC-AGI-2 again doesn't yield a comparable line, since Opus 4.8's tracked score sits at its High tier (72.1) while Gemini's is a flat, untiered 77.1 — different conditions, no matched tier to compare. HMMT Feb 2026 doesn't add a usable line either: both scores (95.45 vs 94.70) carry MathArena's own disclosed contamination-warning, since every model this site tracks was released after the competition happened, so the whole benchmark is graded tainted rather than read as a real gap or a tie. Gemini's remaining case is price: $2/$12 vs $5/$25.

Which is cheaper, Claude Opus 4.8 or Gemini 3.1 Pro Preview?

Gemini 3.1 Pro Preview costs less per output token ($12.00 vs $25.00 per 1M tokens, official listed rates).

Is Claude Opus 4.8 or Gemini 3.1 Pro Preview open weight?

Neither — Claude Opus 4.8 and Gemini 3.1 Pro Preview are both proprietary, so the open-weights question doesn't separate this matchup. What separates it is agent work: Terminal-Bench 2.1 (78.9 vs 65.8) and Toolathlon-Verified (76.2 vs 61.1) are real gaps with independent runs on both sides.

Which is better at agent work, Opus 4.8 or Gemini 3.1 Pro Preview?

Claude Opus 4.8, on the evidence that survives review. Toolathlon-Verified (76.2 vs 61.1) is the cleaner real gap; the Terminal-Bench 2.1 win (78.9 vs 65.8) is weaker because the two rows mix agent scaffolds — Claude Code for Opus, Gemini CLI for Gemini — and vals.ai's single Terminus 2 harness puts them at 71.91 and 70.79, a tie.

Are Claude Opus 4.8 and Gemini 3.1 Pro Preview still current?

Claude Opus 4.8 is computed as a previous version (successor: Claude Opus 5). No same-line successor is tracked for Gemini 3.1 Pro Preview on this site, so its record below stands as the current one for that line.

Why doesn't the Agents' Last Exam comparison count for Opus 4.8 vs Gemini 3.1 Pro Preview?

Because the two rows were run at different effort tiers: Opus 4.8's is the board's Max run while Gemini's is a High run, and more thinking is not a free upgrade. Since 2026-10-02 the comparison is setup-dependent rather than a win, which is why the 27.0 vs 16.4 overall pass rate no longer appears in the verdict.