Skip to content
Bencher

Models · Google (Gemini / DeepMind) · Out sinceReleased 18 Nov 2025

Gemini 3 Pro#10 for writing.#10 for writing, best at its default setting.

Only from the makerNot on OpenRouterBeing retiredDeprecatedClosed (can't be downloaded)Proprietary

Gemini 3 Pro is made by Google (Gemini / DeepMind). Among the models we track it ranks #10 for writing.

Not on OpenRouter as of 2026-10-01 (superseded by Gemini 3.1 Pro preview).

Writing & creativity
60.0 / 100 · #10
Research & analysis
—
Coding
—
Price
Price unknown— / —
Price per 1M (blended)Blended / 1M
—
MemoryContext
—
Longest answerMax output
—
Test resultsResults
21 (21 independent21 indep.)
Out sinceReleased
18 Nov 2025
Made byVendor
Google (Gemini / DeepMind)
UnderstandsInputs
—

Scores are out of 100 for each area, compared with every model we track; “#” is its rank. Context is how much text it can read at once (1M is roughly 700,000 words). Blended price mixes the cost of what you send and what it writes back.

Thinking levels

Reasoning settings

You can't choose how long this one thinks.

No adjustable reasoning setting listed.

Read more

Links

Every test result

Every result

All the numbers,with where they came from.

Every result we've found for this model, with who measured it and a link to where we read it.

All raw rows (vendor and independent kept separate), with setting label, source, date, harness and notes.

All results for this model
NotesNotes
GPQA Diamond91.7%Highreasoning_effort=highIndependent testIndependentVals.ai ↗1 Sep 2026—±1.389 stderr; $0.090/test
GPQA Diamond92.6%Defaultgemini-3-pro-previewIndependent testIndependentEpoch AI ↗19 Nov 2025—Epoch-run GPQA Diamond; stderr 1.7pt; data https://epoch.ai/data/benchmark_data.zip (gpqa_diamond.csv)
Humanity's Last Exam37.5%Defaultgemini-3-pro-previewIndependent testIndependentScale AI SEAL ↗1 Oct 2026—rank 5 (Scale rank accounts for CI); ±1.9; entry added 2025-11-19
LegalBench (Vals)87.0%Highreasoning_effort=highIndependent testIndependentVals.ai ↗29 Sep 2026—vals id google/gemini-3-pro-preview; rank 6/149; ±0.368 stderr; $0.01046/test
LiveCodeBench86.4%DefaultIndependent testIndependentVals.ai ↗1 Sep 2026—±0.982 stderr; $0.127/test
LMArena Search Arena1207Defaultgemini-3-pro-groundingIndependent testIndependentLMArena ↗24 Aug 2026—rank 10 (rank range 8-16); 95% CI 1201.8-1212.9; 37024 votes
LMArena Text - Creative Writing1484Defaultgemini-3-proIndependent testIndependentLMArena ↗30 Sep 2026—rank 7 (rank range 5-20); 95% CI 1476.3-1492.6; 6509 votes; style-controlled
LMArena Text - Multi-Turn1496Defaultgemini-3-proIndependent testIndependentLMArena ↗30 Sep 2026—rank 14 (rank range 7-37); 95% CI 1488.1-1503.6; 6855 votes; style-controlled
LMArena Text - Non-English1478Defaultgemini-3-proIndependent testIndependentLMArena ↗30 Sep 2026—rank 17 (rank range 7-28); 95% CI 1472.7-1482.2; 25023 votes; style-controlled
LMArena Text - Occupational: Legal & Government1502Defaultgemini-3-proIndependent testIndependentLMArena ↗30 Sep 2026—rank 11 (rank range 1-41); 95% CI 1490.6-1512.4; 3037 votes; style-controlled
LMArena Text - Occupational: Writing, Literature & Language1481Defaultgemini-3-proIndependent testIndependentLMArena ↗30 Sep 2026—rank 13 (rank range 7-28); 95% CI 1473.9-1487.7; 9525 votes; style-controlled
LMArena Text (overall)1486Defaultgemini-3-proIndependent testIndependentLMArena ↗30 Sep 2026—text overall, style control on; rank 19 (CI rank 9-28); 95% CI 1482-1489; 41910 votes
MCP Atlas70.3%Defaultgemini-3-pro-previewIndependent testIndependentScale AI SEAL ↗1 Oct 2026—rank 13 (Scale rank accounts for CI); ±2.8; entry added 2025-12-18
METR 50% time horizon3.7 hDefaultIndependent testIndependentMETR ↗8 May 2026METR react agent (Inspect)50% time horizon, METR-Horizon-v1.1; 95% CI 140-379 min; 80% horizon 54.1 min; release 2025-11-18; raw data https://metr.org/assets/benchmark_results_1_1.yaml
SWE-bench Multilingual68.7%DefaultIndependent testIndependentSWE-bench ↗13 Feb 2026mini-SWE-agent 2.0.0a0Official SWE-bench bash-only (mini-SWE-agent) run; data file https://raw.githubusercontent.com/SWE-bench/swe-bench.github.io/master/data/leaderboards.json; leaderboard not updated since 2026-02-26
SWE-Bench Pro (private/commercial set)17.9%DefaultGemini 3 ProIndependent testIndependentScale AI SEAL ↗1 Oct 2026—rank 9 (Scale rank accounts for CI); ±4.78; entry added 2025-09-19
SWE-Bench Pro (public, v1)43.3%Defaultgemini-3-pro-previewIndependent testIndependentScale AI SEAL ↗1 Oct 2026—rank 5 (Scale rank accounts for CI); ±3.6; entry added 2025-11-26
SWE-bench Verified (bash-only, mini-SWE-agent)69.6%HighIndependent testIndependentSWE-bench ↗26 Feb 2026mini-SWE-agent 2.0.0Official SWE-bench bash-only (mini-SWE-agent) run; data file https://raw.githubusercontent.com/SWE-bench/swe-bench.github.io/master/data/leaderboards.json; leaderboard not updated since 2026-02-26
Terminal-Bench 2.173.9%DefaultIndependent testIndependentSnorkel AI / Terminal-Bench ↗1 Oct 2026Terminus 2TB 2.1 (89 tasks, archived); effort not listed
Terminal-Bench 2.165.8%DefaultIndependent testIndependentSnorkel AI / Terminal-Bench ↗1 Oct 2026Gemini CLITB 2.1 (89 tasks, archived); effort not listed
Vectara Hallucination Leaderboard (HHEM)13.6%Defaultgoogle/gemini-3-pro-previewIndependent testIndependentVectara ↗22 Sep 2026—factual consistency 86.4 %; answer rate 99.4 %; avg summary 101.9 words; HHEM-2.3 judge; effort not stated (API default)