Inference
EN

Gemini 3.1 Pro Preview vs Mistral Large 4

Change models
Gemini 3.1 Pro PreviewGoogle
Mistral Large 4Mistral
IndexAll categories, out of 100.
61.5 ±3.6 best
60.7 ±4.1
GapIndex points below the top.
Top
−0.8 · no clear gap
CoverageIndex weight that has results.
95% best
85%
SpeedOutput tokens per second.
77/s
113/s best
EloLMArena rating and rank.
1480 best
1414
ReleasedMonth of release.
2026-02
2026-10
Input / 1MUS dollars per 1M input tokens.
$2.00
$0.70 best
Output / 1MUS dollars per 1M output tokens.
$12.00
$2.10 best
Batch output / 1MBatch price per 1M output tokens.
$6.00
—
1K chats2K tokens in, 500 out, each.
$10.00
$2.45 best
1K documents20K tokens in, 1K out, each.
$52.00
$16.10 best
1K agent tasks200K tokens in, 20K out, each.
$640.00
$182.00 best
ContextMaximum tokens in one request.
1.05M best
1M
AgenticMulti-step tasks with tools.
52.5
66.3 best
CodingCode writing and repair.
58.8
62.7 best
ReasoningLogic problems and puzzles.
69.8 best
59.2
MultimodalTasks with images and text.
69.9
—
KnowledgeFacts and expert knowledge.
65.0 best
43.7
MultilingualTasks in many languages.
—
—
InstructionTasks with strict prompt rules.
86.6 best
57.7
MathMath problems.
60.4
61.6 best

Benchmarks

29 shared
Only Gemini 3.1 Pro Preview: 27Only Mistral Large 4: 31
Gemini 3.1 Pro Preview
Mistral Large 4
Vibe Code Bench v1.1Coding
32.0%80 / 123
78.4% best35 / 123
MedScribeHealthcare
76.1%77 / 109
80.4% best57 / 109
MedCodeHealthcare
59.1% best2 / 106
40.7%61 / 106
Legal Research BenchLegal
20.7%67 / 90
31.7% best51 / 90
Tax Agent BenchFinance
9.3%68 / 82
30.4% best33 / 82
LiveBench Data AnalysisReasoning
78.5% best23 / 80
76.5%33 / 80
LiveBench ReasoningReasoning
84.0% best42 / 80
83.9%43 / 80
Finance Agent (v2)Finance
43.0%59 / 79
54.7% best26 / 79
Code MigrationCoding
17.3%54 / 77
30.6% best41 / 77
EMBFinance
52.6%54 / 75
56.2% best47 / 75
LiveBench Agentic CodingAgentic
44.1%54 / 66
57.2% best22 / 66
LiveBench CodingCoding
76.5%43 / 66
77.2% best37 / 66
LiveBench Instruction FollowingInstruction
79.1% best3 / 66
64.8%42 / 66
LiveBench LanguageKnowledge
85.4% best12 / 66
49.6%65 / 66
LiveBench MathematicsMath
91.0%28 / 66
93.6% best18 / 66
Terminal-Bench 4.0Agentic
2.5%54 / 61
27.8% best26 / 61
Agent Arena steerabilityAgentic
−8.243 / 54
−6.4 best41 / 54
Agent Arena task outcomeAgentic
−11.347 / 54
−4.7 best39 / 54
ProofBench v1.1Math
26.0% best40 / 54
10.0%49 / 54
LiveBench Coding 2.0Coding
66.7% best41 / 52
60.0%47 / 52