Inference
EN

DeepSeek V4.1 Flash 1017 vs Gemini 4 Carbon

Change models
DeepSeek V4.1 Flash 1017DeepSeekFastest
Gemini 4 CarbonGoogle
IndexAll categories, out of 100.
68.5 ±2.6
82.7 ±5.1 best
GapIndex points below the top.
−14.2
Top
CoverageIndex weight that has results.
95% best
80%
SpeedOutput tokens per second.
261/s best
31/s
EloLMArena rating and rank.
1463
1538 best
ReleasedMonth of release.
2026-10
2026-10
Input / 1MUS dollars per 1M input tokens.
$0.022 best
$2.00
Output / 1MUS dollars per 1M output tokens.
$0.40 best
$10.00
Batch output / 1MBatch price per 1M output tokens.
$0.20 best
$5.00
1K chats2K tokens in, 500 out, each.
$0.244 best
$9.00
1K documents20K tokens in, 1K out, each.
$0.84 best
$50.00
1K agent tasks200K tokens in, 20K out, each.
$12.40 best
$600.00
ContextMaximum tokens in one request.
1.05M
1.05M
AgenticMulti-step tasks with tools.
72.2
83.2 best
CodingCode writing and repair.
69.8
80.5 best
ReasoningLogic problems and puzzles.
65.0
83.9 best
MultimodalTasks with images and text.
62.2
—
KnowledgeFacts and expert knowledge.
64.2
79.5 best
MultilingualTasks in many languages.
—
—
InstructionTasks with strict prompt rules.
68.3
—
MathMath problems.
65.4
77.2 best

Benchmarks

26 shared
Only DeepSeek V4.1 Flash 1017: 68Only Gemini 4 Carbon: 13
DeepSeek V4.1 Flash 1017
Gemini 4 Carbon
Artificial Analysis Long Context ReasoningReasoning
84.0% best8 / 217
79.3%52 / 217
GPQA diamond · Epoch AIKnowledge
89.8%42 / 217
96.5% best1 / 217
OTIS Mock AIME 2024-2025Math
98.3%28 / 200
100.0% best1 / 200
Artificial Analysis Omniscience AccuracyKnowledge
46.4%33 / 164
58.3% best9 / 164
ARC-AGI-2 (semi-private)Reasoning
72.9%25 / 131
92.1% best5 / 131
ARC-AGI-1 (semi-private)Reasoning
94.5%21 / 126
97.5% best5 / 126
Vibe Code Bench v1.1Coding
84.7%18 / 123
89.1% best9 / 123
ARC-AGI-2 (public eval)Reasoning
81.7%23 / 122
97.9% best3 / 122
ARC-AGI-1 (public eval)Reasoning
98.0%11 / 116
98.5% best6 / 116
SimpleQA VerifiedKnowledge
42.9%47 / 91
77.7% best2 / 91
Chess PuzzlesReasoning
48.0%19 / 90
78.0% best2 / 90
FrontierMath-Tiers-1-3-v2-PrivateMath
67.4%33 / 90
90.5% best6 / 90
Legal Research BenchLegal
41.3%30 / 90
55.8% best1 / 90
Tax Agent BenchFinance
26.6%41 / 82
51.8% best1 / 82
Mystery Game PuzzlesReasoning
43.0%17 / 80
73.0% best5 / 80
FrontierMath-Tier-4-v2-PrivateMath
26.8%39 / 72
95.1% best4 / 72
Terminal-Bench 4.0Agentic
19.7%34 / 61
63.1% best4 / 61
Agent Arena steerabilityAgentic
4.018 / 54
19.1 best2 / 54
Agent Arena task outcomeAgentic
2.325 / 54
18.4 best4 / 54
Furniture AssemblyReasoning
34.2%24 / 37
77.5% best6 / 37