Pass IndexThe State of AISign in

gemini-3-pro-preview

Gemini 3.1 Pro is the next iteration in the Gemini 3 series of models, a suite of highly capable, natively multimodal reasoning models.

text → text · made by Google

$2→$12per Mtok in / out
Databricks · 3 sellers

Sold by 8 ways

SellerLaneRate
Google Vertex AI cloudbatch$1per Mtok in$6per Mtok outraw.githubusercontent.com · read 2026-09-16
Databricks cloudstandard$2per Mtok in$12per Mtok out$0.2per Mtok cached$2.5per Mtok cache writemodels.dev · read 2026-09-16
Google Vertex AI cloudstandard$2per Mtok in$12per Mtok out$0.2per Mtok cachedraw.githubusercontent.com · read 2026-09-16
Google Vertex AI cloudpriority$3.6per Mtok in$21.6per Mtok out$0.36per Mtok cachedraw.githubusercontent.com · read 2026-09-16
Google Vertex AI cloudlong-context$4per Mtok in$18per Mtok out$0.4per Mtok cachedraw.githubusercontent.com · read 2026-09-16
Requesty 2 ways$0.9per Mtok in$5.4per Mtok out$0.18per Mtok cached
Requesty aggregatornot on the seller's list todaygoogle flex$0.9per Mtok in$5.4per Mtok out$0.18per Mtok cachedrouter.requesty.ai · read 2026-08-25
Requesty aggregatorgoogle$1per Mtok in$6per Mtok out$0.2per Mtok cachedrouter.requesty.ai · read 2026-09-16
Requesty aggregatorvertex$2per Mtok in$12per Mtok out$0.2per Mtok cachedrouter.requesty.ai · read 2026-08-28

Measured 51 standings

PlaceBoardMetricScore
1stof 36Balrog — Epoch AIAverage progress0.581
1stof 10Humanity's Last ExamAccuracy (%)38.3
1stof 38Vpct — Epoch AICorrect0.91
2ndof 262MMLU-ProOverall (accuracy) (11/25)90.1
3rdof 109Berkeley Function-Calling Leaderboard (BFCL) V4Overall Acc (prompt)72.51
4thof 18Algotune — Epoch AIScore1.83
4thof 180SWE-bench Verified (Bash Only toggle off — all agents)% Resolved77.4
5thof 11Gdpval — Epoch AIWin Rate (%)0.403
5thof 51Humanity's Last Exam (Epoch AI replication)Accuracy37.52
5thof 25SWE-Bench Pro (Public Dataset)Resolve Rate43.3
5thof 8τ²-bench Leaderboard (Overall: Retail · Airline · Telecom)Pass^182.5
6thof 47SWE-bench Verified (default "Bash Only" view, agent = mini-SWE-agent)% Resolved74.2
7thof 109Berkeley Function-Calling Leaderboard (BFCL) V4Overall Acc (fc)68.14
7thof 102Simplebench — Epoch AIScore (AVG@5)0.764
8thof 61AIME 2025Accuracy (± 95% CI) (preview)95
8thof 46Enigma eval — Epoch AIAccuracy18.24
10thof 34LMArena · SearchScore (Elo)1207
10thof 50Metr time horizons — Epoch AIaverage_score0.71
11thof 17terminal-bench@2.1Accuracy73.9
12thof 81Forecastbench — Epoch AIOverall score61.2
12thof 152LMArena · VisionScore (Elo)1289
12thof 13Rli — Epoch AIScore1.25
15thof 72Frontiermath tier 4 — Epoch AImean_score18.75
15thof 38Gso — Epoch AIScore OPT@10.186
15thof 47SWE-bench Verified (default "Bash Only" view, agent = mini-SWE-agent)% Resolved (high)69.6
16thof 402LMArena · TextScore (Elo)1485
17thof 23Cl bench — Epoch AIOverall0.158
17thof 101FrontierMath (Tiers 1-3)mean_score37.6
19thof 60Vending bench 2 — Epoch AIScore5478.158
19thof 60Vending-Bench 2 — Current leaderboardMoney Balance (average across 5 runs)5478.16
21stof 313GPQA Diamondmean_score92.614
21stof 28τ³-Banking LeaderboardPass^118
22ndof 35Swe bench verified — Epoch AImean_score72.934
24thof 142terminal-bench@2.0Accuracy69.4
27thof 65Apex agents — Epoch AIPass@1 score0.315
27thof 202Terminalbench — Epoch AIAccuracy mean0.694
28thof 41Deepresearchbench — Epoch AIAverage score (low)0.463
32ndof 112Ale bench — Epoch AIPerformance1176.75
32ndof 170Weirdml — Epoch AIAccuracy69.93
36thof 64Proofbench — Epoch AIAccuracy20
45thof 222Chess puzzles — Epoch AImean_score31
56thof 290OTIS Mock AIME 2024-2025mean_score91.389
61stof 128LMArena · WebDevScore (Elo)1439
61stof 121Webdev arena — Epoch AIArena Score1437.8
76thof 177Critpt — Epoch AIAccuracy6.9
86thof 105Vectara Hallucination LeaderboardHallucination Rate13.6
92ndof 232ARC-AGI-2 (Semi-Private)Score (%)31.11
99thof 227Arc agi 2 — Epoch AIScore31.11
101stof 230ARC-AGI-1 (Semi-Private)Score (%)75
109thof 244Arc agi — Epoch AIScore75
173rdof 656Epoch capabilities index — Epoch AIECI Score152.94

About

gemini-3-pro-preview — a text model from Google, sold by 3 companies from $2 in and $12 out per million tokens, placed 1st of 38 on Vpct — Epoch AI.

It takes text and returns text, with a context window of 1,048,576 tokens. It was published in October 2025, trained on material up to January 2025. Its sellers say it can reason step by step and call a tool. The catalogue files it under reasoning, code and agents. Three companies sell it. The cheapest is $2 in and $12 out per million tokens at Databricks. Beside the standard rate there are batch, separately routed and faster lanes. It has been measured on 49 boards, and stands best at 1st of 38 on Vpct — Epoch AI.

Every current figure

Maker
Google
Register
model
Takes
text
Returns
text
Context
1,048,576 tokens
Longest answer
65,536 tokens
Published
October 2025
Knowledge to
January 2025
Licence
not read
Sellers
3
Maker's own price
not read
Price
$2 in and $12 out per million tokens — Databricks
Boards
49
Best place
1st of 38 — Vpct — Epoch AI

Known as 14 names

Gemini 3 Pro (preview)Gemini-3-Pro-Preview (FC)Gemini-3-Pro-Preview (Prompt)Gemini 3 Pro PreviewGemini 3 Progemini-3-pro-groundinggemini-3-proGemini-3-Pro(11/25)Gemini 3 Pro Preview (2025-11-18)Gemini 3 Pro highlive-SWE-agent + Gemini 3 Pro Preview (2025-11-18)google/gemini-3-pro-previewgoogle/gemini-3-pro-preview:flexvertex/gemini-3-pro-preview