Inference
EN

Claude Opus 5 vs Gemini 4 Carbon

Change models
Claude Opus 5Anthropic
Gemini 4 CarbonGoogle
IndexAll categories, out of 100.
73.8 ±1.9
82.7 ±5.1 best
GapIndex points below the top.
−8.9
Top
CoverageIndex weight that has results.
100% best
80%
SpeedOutput tokens per second.
48/s best
31/s
EloLMArena rating and rank.
1502
1538 best
ReleasedMonth of release.
2026-07
2026-10
Input / 1MUS dollars per 1M input tokens.
$5.00
$2.00 best
Output / 1MUS dollars per 1M output tokens.
$25.00
$10.00 best
Batch output / 1MBatch price per 1M output tokens.
$12.50
$5.00 best
1K chats2K tokens in, 500 out, each.
$22.50
$9.00 best
1K documents20K tokens in, 1K out, each.
$125.00
$50.00 best
1K agent tasks200K tokens in, 20K out, each.
$1,500.00
$600.00 best
ContextMaximum tokens in one request.
1M
1.05M best
AgenticMulti-step tasks with tools.
71.6
83.2 best
CodingCode writing and repair.
74.8
80.5 best
ReasoningLogic problems and puzzles.
75.4
83.9 best
MultimodalTasks with images and text.
72.5
—
KnowledgeFacts and expert knowledge.
73.1
79.5 best
MultilingualTasks in many languages.
63.6
—
InstructionTasks with strict prompt rules.
55.6
—
MathMath problems.
74.8
77.2 best

Benchmarks

28 shared
Only Claude Opus 5: 122Only Gemini 4 Carbon: 11
Claude Opus 5
Gemini 4 Carbon
Artificial Analysis Long Context ReasoningReasoning
79.3%52 / 217
79.3%52 / 217
GPQA diamond · Epoch AIKnowledge
93.9%14 / 217
96.5% best1 / 217
OTIS Mock AIME 2024-2025Math
98.9%22 / 200
100.0% best1 / 200
Artificial Analysis Omniscience AccuracyKnowledge
60.9% best7 / 164
58.3%9 / 164
ARC-AGI-2 (semi-private)Reasoning
90.4%8 / 131
92.1% best5 / 131
ARC-AGI-1 (semi-private)Reasoning
97.5%5 / 126
97.5%5 / 126
Vibe Code Bench v1.1Coding
88.4%11 / 123
89.1% best9 / 123
ARC-AGI-2 (public eval)Reasoning
97.1%6 / 122
97.9% best3 / 122
ARC-AGI-1 (public eval)Reasoning
99.0% best2 / 116
98.5%6 / 116
SimpleQA VerifiedKnowledge
59.9%19 / 91
77.7% best2 / 91
Chess PuzzlesReasoning
43.0%31 / 90
78.0% best2 / 90
FrontierMath-Tiers-1-3-v2-PrivateMath
85.6%15 / 90
90.5% best6 / 90
Legal Research BenchLegal
55.3%2 / 90
55.8% best1 / 90
Tax Agent BenchFinance
47.1%5 / 82
51.8% best1 / 82
Mystery Game PuzzlesReasoning
59.0%9 / 80
73.0% best5 / 80
FrontierMath-Tier-4-v2-PrivateMath
73.2%15 / 72
95.1% best4 / 72
Terminal-Bench 4.0Agentic
53.5%10 / 61
63.1% best4 / 61
Agent Arena steerabilityAgentic
4.417 / 54
19.1 best2 / 54
Agent Arena task outcomeAgentic
10.210 / 54
18.4 best4 / 54
Furniture AssemblyReasoning
60.8%10 / 37
77.5% best6 / 37