Gemini 3.1 Pro vs Qwen3.5 397B A17B
Wins 7 of 8 areas
Coding · Agents · Reasoning · Facts · Images and charts · Math · Long documents
Wins 0 of 8 areas
—
Gemini 3.1 Pro is the stronger all-rounder.Qwen3.5 397B A17B is cheaper.
Both rank among the ten best models we track in following instructions.
Scores updated · 56 tests both models report · How we compare
Where each one wins
Tests won in each of the eight areas we test. Each piece is one test, so a longer bar means more evidence; grey means the two scored within a point of each other.
- CodingWriting and fixing software80Gemini 3.1 Pro8 of 8 tests
- MathCompetition and research-level math40Gemini 3.1 Pro4 of 5 tests · 1 tie
- AgentsCarrying out multi-step tasks on its own41Gemini 3.1 Pro4 of 6 tests · 1 tie
- ReasoningHard problems that need careful thinking30Gemini 3.1 Pro3 of 3 tests
- FactsGetting facts right instead of making them up30Gemini 3.1 Pro3 of 3 tests
- Images and chartsUnderstanding pictures, charts and video20Gemini 3.1 Pro2 of 2 tests
- Long documentsFinding answers in very long texts10Gemini 3.1 Pro1 of 1 test
- Following instructionsDoing exactly what it is asked11Even1 each
Long documents rests on a single test.
What it costs
Prices per million tokens, roughly 750,000 words. The bars show the cost of a million tokens read plus a million written.
Qwen3.5 397B A17B costs 70% less for the same work.
The biggest differences
The tests each model wins by the widest margin, up to three each. Scores are out of 100.
Where Gemini 3.1 Pro pulls ahead
- Short factual questions, answered correctlySimpleQA Verified+47.5points ahead
- Proof problems from the US Math OlympiadUSAMO 2026+38.1points ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+38points ahead
Where Qwen3.5 397B A17B pulls ahead
- Finds the root cause of IT system incidentsAA IT-Bench SRE+3.8points ahead
- Follows unfamiliar, precisely checkable instructionsIFBench+1.7points ahead
Every test, side by side
All 56 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingGemini 3.1 Pro
- Terminal-Bench 2.1Gemini 3.1 Pro by 22.573.851.3+22.5
- SciCodeGemini 3.1 Pro by 13.958.744.8+13.9
- Terminal-Bench HardGemini 3.1 Pro by 12.953.840.9+12.9
- LiveCodeBench v6Gemini 3.1 Pro by 8.191.783.6+8.1
- SWE-bench MultilingualGemini 3.1 Pro by 7.676.969.3+7.6
- LMArena · WebDevGemini 3.1 Pro by 47 rating points14461399+47 rating
- SWE-bench VerifiedGemini 3.1 Pro by 4.280.676.4+4.2
- SWE-bench ProGemini 3.1 Pro by 3.354.250.9+3.3
AgentsGemini 3.1 Pro
- BrowseCompGemini 3.1 Pro by 16.985.969+16.9
- AA ApexAgentsGemini 3.1 Pro by 16.73215.3+16.7
- τ-Bench V3 · BankingGemini 3.1 Pro by 821.413.4+8
- Terminal-Bench 4.0Gemini 3.1 Pro by 440+4
- AA IT-Bench SREQwen3.5 397B A17B by 3.830.334.1+3.8
- GDPValtie14.714.8tie
ReasoningGemini 3.1 Pro
- Humanity's Last ExamGemini 3.1 Pro by 184729+18
- CritPtGemini 3.1 Pro by 1617.71.7+16
- GPQA DiamondGemini 3.1 Pro by 4.894.189.3+4.8
FactsGemini 3.1 Pro
- SimpleQA VerifiedGemini 3.1 Pro by 47.573.526+47.5
- AA-Omniscience · Non-hallucinationGemini 3.1 Pro by 3849.111.1+38
- AA-Omniscience · AccuracyGemini 3.1 Pro by 24.154.930.8+24.1
Images and chartsGemini 3.1 Pro
- MMMU-ProGemini 3.1 Pro by 5.182.477.3+5.1
- LMArena · VisionGemini 3.1 Pro by 32 rating points12961264+32 rating
MathGemini 3.1 Pro
- USAMO 2026Gemini 3.1 Pro by 38.174.436.3+38.1
- FrontierMath Tiers 1-3 (v2)Gemini 3.1 Pro by 28.559.631.2+28.5
- AIME 2026Gemini 3.1 Pro by 798.391.3+7
- HMMT Feb. 2026Gemini 3.1 Pro by 6.894.787.9+6.8
- IMOAnswerBenchtie8180.9tie
Following instructionsEven
- Multi-ChallengeGemini 3.1 Pro by 3.871.467.6+3.8
- IFBenchQwen3.5 397B A17B by 1.777.178.8+1.7
Other results26 tests, not counted
Tests outside the eight areas. They are not counted above: several are summary scores built from other tests, or the same test under another name.
- AA-OmniscienceGemini 3.1 Pro by 62.631.9-30.8+62.6
- Artificial Analysis Coding IndexGemini 3.1 Pro by 20.668.848.2+20.6
- Terminal-Bench 2.0Gemini 3.1 Pro by 1668.552.5+16
- t2-benchGemini 3.1 Pro by 12.699.386.7+12.6
- AA IntelligenceGemini 3.1 Pro by 11.329.718.4+11.3
- ToolathlonGemini 3.1 Pro by 10.548.838.3+10.5
- MCPMarkGemini 3.1 Pro by 9.855.946.1+9.8
- AgentWorldBench - MCPQwen3.5 397B A17B by 9.259.168.3+9.2
- AgentWorldBench - AndroidGemini 3.1 Pro by 6.561.454.9+6.5
- AgentWorldBench - OSGemini 3.1 Pro by 666.960.9+6
- AgentWorldBench - SWEQwen3.5 397B A17B by 5.359.164.4+5.3
- AgentWorldBench - WebGemini 3.1 Pro by 4.252.848.5+4.2
- MMMLUGemini 3.1 Pro by 4.192.688.5+4.1
- SWEBench Pro PublicGemini 3.1 Pro by 3.354.250.9+3.3
- MMLU-ProGemini 3.1 Pro by 3.29187.8+3.2
- HLE (with tools)Gemini 3.1 Pro by 3.151.448.3+3.1
- τ³-Bench BankingGemini 3.1 Pro by 3.116.513.4+3.1
- AgentWorldBench - TerminalQwen3.5 397B A17B by 2.852.555.3+2.8
- HMMT Nov. 2025Gemini 3.1 Pro by 2.194.892.7+2.1
- StrongREJECTQwen3.5 397B A17B by 1.49899.4+1.4
- NL2RepoGemini 3.1 Pro by 1.233.432.2+1.2
- AgentWorldBench - Searchtie30.230.8tie
- AA Agentic Indextie10.310.6tie
- AgentWorldBench - Overalltie54.654.7tie
- GDPval-AA v2tie962962tie
- τ²-Bench Telecom (AA run)tie95.695.6tie
Questions people ask
Which is better, Gemini 3.1 Pro or Qwen3.5 397B A17B?
Gemini 3.1 Pro wins seven of the eight areas we test: coding, agents, reasoning, facts, images and charts, math and long documents. Qwen3.5 397B A17B wins none, but costs 70% less. They are level on following instructions.
Which is better for coding?
Gemini 3.1 Pro. It wins 8 of the 8 coding tests both models report; Qwen3.5 397B A17B wins none.
Which is cheaper?
Gemini 3.1 Pro costs $2.00 per million input tokens and $12.00 per million output tokens; Qwen3.5 397B A17B costs $0.60 and $3.60. That makes Qwen3.5 397B A17B about 70% cheaper for the same work.
How do you compare the two?
We use the 56 benchmark tests both models have published scores on. The verdict counts the 30 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 26 are listed but not counted, because several are summary scores or repeat a test. Each score is the one shown on the model's own page.