Inference
EN

Gemini 3 Pro vs Kimi K2.7 Code

Change models
Gemini 3 ProGoogle
Kimi K2.7 CodeMoonshot AI
IndexAll categories, out of 100.
57.9 ±2.7 best
57.7 ±4.5
GapIndex points below the top.
Top
−0.2 · no clear gap
CoverageIndex weight that has results.
100% best
85%
SpeedOutput tokens per second.
109/s best
35/s
EloLMArena rating and rank.
1480
—
ReleasedMonth of release.
2025-11
2026-06
Input / 1MUS dollars per 1M input tokens.
$2.00
$0.60 best
Output / 1MUS dollars per 1M output tokens.
$12.00
$3.50 best
1K chats2K tokens in, 500 out, each.
$10.00
$2.95 best
1K documents20K tokens in, 1K out, each.
$52.00
$15.50 best
1K agent tasks200K tokens in, 20K out, each.
$640.00
$190.00 best
ContextMaximum tokens in one request.
2M best
262K
AgenticMulti-step tasks with tools.
56.2
58.4 best
CodingCode writing and repair.
56.4
60.4 best
ReasoningLogic problems and puzzles.
57.2 best
54.6
MultimodalTasks with images and text.
54.8
—
KnowledgeFacts and expert knowledge.
66.1 best
58.8
MultilingualTasks in many languages.
60.1
—
InstructionTasks with strict prompt rules.
54.8 best
44.2
MathMath problems.
58.2 best
57.0

Benchmarks

11 shared
Only Gemini 3 Pro: 64Only Kimi K2.7 Code: 26
Gemini 3 Pro
Kimi K2.7 Code
Artificial Analysis Humanity's Last ExamKnowledge
39.7% best48 / 222
35.0%63 / 222
Artificial Analysis Long Context ReasoningReasoning
76.0%88 / 217
79.3% best52 / 217
GPQA diamond · Epoch AIKnowledge
92.6% best27 / 217
87.9%59 / 217
Critical Physics TasksReasoning
9.1%70 / 214
10.0% best65 / 214
OTIS Mock AIME 2024-2025Math
91.4%56 / 200
95.6% best40 / 200
Artificial Analysis GPQA DiamondKnowledge
90.8% best30 / 199
89.6%41 / 199
τ²-Bench Tool-Agent-User EvaluationAgentic
87.1%48 / 149
90.1% best40 / 149
Artificial Analysis IFBenchInstruction
70.4% best42 / 146
63.1%60 / 146
LiveCodeBenchCoding
86.4% best23 / 142
82.0%57 / 142
Vibe Code Bench v1.1Coding
14.3%99 / 123
47.2% best73 / 123
SWE-benchCoding
76.4%37 / 87
78.2% best30 / 87