Inference
EN

Gemini 4 Argon vs Kimi K3.1 Pro

Change models
Gemini 4 ArgonGoogle
Kimi K3.1 ProMoonshot AISmartest open
IndexAll categories, out of 100.
78.5 ±4.3 best
75.7 ±4.7
GapIndex points below the top.
Top
−2.8 · no clear gap
CoverageIndex weight that has results.
80%
80%
SpeedOutput tokens per second.
28/s
43/s best
EloLMArena rating and rank.
1530
—
ReleasedMonth of release.
2026-09
2026-10
Input / 1MUS dollars per 1M input tokens.
$2.00
$2.00
Output / 1MUS dollars per 1M output tokens.
$10.00
$10.00
Batch output / 1MBatch price per 1M output tokens.
$5.00 best
$6.00
1K chats2K tokens in, 500 out, each.
$9.00
$9.00
1K documents20K tokens in, 1K out, each.
$50.00
$50.00
1K agent tasks200K tokens in, 20K out, each.
$600.00
$600.00
ContextMaximum tokens in one request.
1.05M
1.05M
AgenticMulti-step tasks with tools.
77.3 best
76.7
CodingCode writing and repair.
76.7 best
71.9
ReasoningLogic problems and puzzles.
79.8 best
77.7
MultimodalTasks with images and text.
—
72.0
KnowledgeFacts and expert knowledge.
76.5 best
71.0
MultilingualTasks in many languages.
—
—
InstructionTasks with strict prompt rules.
—
—
MathMath problems.
75.9
—

Benchmarks

17 shared
Only Gemini 4 Argon: 69Only Kimi K3.1 Pro: 46
Gemini 4 Argon
Kimi K3.1 Pro
Artificial Analysis Humanity's Last ExamKnowledge
57.1% best4 / 222
52.5%11 / 222
Artificial Analysis Long Context ReasoningReasoning
79.7%45 / 217
86.3% best3 / 217
Critical Physics TasksReasoning
27.1% best17 / 214
24.6%22 / 214
Artificial Analysis Omniscience AccuracyKnowledge
48.3%26 / 164
48.9% best23 / 164
GDPval-AA normalizedAgentic
56.2%14 / 133
67.3% best3 / 133
Artificial Analysis SciCodeCoding
61.8% best5 / 114
60.8%9 / 114
DeepSWEAgentic
77.9% best3 / 59
75.5%5 / 59
MazeBench: visual spatial reasoning in a 3D open worldAgentic
12.0% best7 / 48
2.0%15 / 48
FrontierCode 2.0Coding
26.3% best14 / 27
19.6%22 / 27
Artificial Analysis AnalystAgentAgentic
48.8% best12 / 26
38.8%23 / 26
Bug Hunt BenchCoding
40.7% best9 / 25
36.0%11 / 25
OSWorld 2.0Agentic
69.2%4 / 25
79.6% best1 / 25
Artificial Analysis AutomationBenchAgentic
77.5% best2 / 23
73.3%4 / 23
ExploitGymAgentic
39.5% best4 / 23
20.7%13 / 23
Agents' Last ExamAgentic
39.5%9 / 21
43.2% best6 / 21
Multimodal Knowledge Agent BenchmarkAgentic
51.3% best5 / 19
43.9%9 / 19
CWE-bench v2Agentic
65.0% best3 / 11
45.3%10 / 11