Inference
EN
xAI

Grok 4.6

Grok 4.6 is a reasoning model from xAI. 54 benchmarks count toward its index, in 6 categories.

  • Closed weights
  • Reasoning
  • Input: text · image · file
  • Released 2026-08
RankedEnough results for an index.
IndexAll categories, out of 100.67.8±3.7Median 51.0
CoverageIndex weight that has results.85%
SpeedOutput tokens per second.56/sMedian 67/s
First tokenSeconds to the first output token.2.8 sMedian 1.3 s
Input / 1MUS dollars per 1M input tokens.$2.00batch $1.60Median $0.60
Output / 1MUS dollars per 1M output tokens.$6.00batch $4.80 Batch price per 1M output tokens.Median $2.55
ContextMaximum tokens in one request.500KMedian 262K
EloLMArena rating and rank.1430±5 #96Median 1422

Details

From OpenRouter
API IDThe model ID on OpenRouter.
x-ai/grok-4.6
Max outputMaximum output tokens per request.
450K
Knowledge cutoffLast date of training data.
—
ToolsTool calls on OpenRouter.
Yes
JSON outputOutput follows a given JSON schema.
Yes
Cache read / 1MUS dollars per 1M cached tokens.
$0.50
WeightsLink to the published weights.
—
RetiresDate OpenRouter removes the model.
—

Where to run

APIs that serve this model.
6 providers
ProviderTypeMaker, cloud, host or router.InputOutputContext
Amazon BedrockCloud$2.00$6.00500K
Google Vertex AICloud$2.00$6.00524K
Microsoft AzureCloud$2.00$6.00200K
OpenRouterRouter$2.00$6.00500K
Oracle CloudCloud$2.00$6.00500K
xAIMaker$2.00$6.00500K

US dollars per 1M tokens, list price. From models.dev (opens in a new tab) (MIT).

Capabilities

Index per category, out of 100.
Out of 100
AgenticMulti-step tasks with tools.16 counted68.432 / 218
CodingCode writing and repair.16 counted67.728 / 209
ReasoningLogic problems and puzzles.10 counted66.142 / 227
MultimodalTasks with images and text.None counted——
KnowledgeFacts and expert knowledge.6 counted65.741 / 268
MultilingualTasks in many languages.None counted——
InstructionTasks with strict prompt rules.1 counted72.0—
MathMath problems.5 counted65.329 / 168

Results

54 counted
BenchmarkThe benchmark name.ResultThe score from the source.PlacePlace among results on this benchmark.
Artificial Analysis GPQA DiamondKnowledge94.9%3 / 199
GPQA Diamond · Vals AIKnowledge94.7% ±1.13 / 135
Artificial Analysis Tau3-BankingAgentic50.7%1 / 22
SWE-benchCoding95.6% ±0.94 / 87
LiveCodeBenchCoding88.2% ±0.97 / 142
MMLU ProKnowledge89.4% ±0.311 / 135
OTIS Mock AIME 2024-2025Math99.2% ±0.519 / 200
GPQA diamond · Epoch AIKnowledge93.2% ±1.521 / 217
LiveBench ReasoningReasoning91.2%8 / 80
cursorBench32Coding70.8%2 / 18
GDPval-AA normalizedAgentic55.9%16 / 133
Artificial Analysis Terminal-Bench v2.1Agentic88.4%3 / 21
Artificial Analysis Humanity's Last ExamKnowledge42.9%34 / 222
Artificial Analysis Omniscience AccuracyKnowledge48.2%27 / 164
LiveBench ResearchAgentic84.5%9 / 52
Terminal-Bench 2.1 · Vals AIAgentic78.3% ±2.113 / 75
Artificial Analysis Long Context ReasoningReasoning80.3%40 / 217
Critical Physics TasksReasoning17.1%42 / 214
Artificial Analysis SciCodeCoding56.5%24 / 114
ARC-AGI-2 (semi-private)Reasoning67.1%31 / 131
LiveBench LanguageKnowledge83.7%17 / 66
Code MigrationCoding44.6% ±4.520 / 77
LiveBench DesignAgentic79.8%14 / 52
VulcanBench v3Coding87.0%4 / 14
Agent Arena command recoveryAgentic5.715 / 50
LiveBench Instruction FollowingInstruction71.9%20 / 66
APEX-AgentsAgentic57.5%4 / 13
Terminal-Bench 3.0Agentic26.5%4 / 13
Mystery Game PuzzlesReasoning34.0% ±4.825 / 80
LiveBench Coding 2.0Coding89.1%17 / 52
Vibe Code Bench v1.1Coding76.2% ±3.841 / 123
MazeBench: visual spatial reasoning in a 3D open worldAgentic1.0%16 / 48
SimpleQA VerifiedKnowledge48.9% ±1.632 / 91
ARC-AGI-1 (semi-private)Reasoning87.0%45 / 126
LiveBench MathematicsMath92.6%24 / 66
LiveBench Agentic CodingAgentic57.0%24 / 66
SkillsBenchCoding55.8% ±4.813 / 35
FrontierMath-Tiers-1-3-v2-PrivateMath66.0% ±2.835 / 90
LiveBench Long AgenticAgentic40.7%20 / 51
ARC-AGI-3 (semi-private)Reasoning2.1%11 / 26
FrontierMath-Tier-4-v2-PrivateMath31.7% ±7.432 / 72
LiveBench Data AnalysisReasoning75.3%39 / 80
LiveBench Browser UseAgentic64.0%25 / 49
EBR-benchReasoning30.5% ±1.616 / 31
Chess PuzzlesReasoning33.0% ±4.748 / 90
MirrorCodeCoding41.1% ±11.021 / 39
Furniture AssemblyReasoning40.0% ±6.320 / 37
DeepSWEAgentic65.9%33 / 59
CWE-bench v1Agentic57.0%7 / 12
ProofBench v1.1Math51.0% ±5.032 / 54
Bug Hunt BenchCoding27.0%15 / 25
Terminal-Bench 4.0Agentic17.2% ±1.337 / 61
IOICoding47.6% ±2.128 / 45
Agent Arena steerabilityAgentic−1.834 / 54
LiveBench CodingCoding76.8%42 / 66
Agent Arena task outcomeAgentic−2.636 / 54
Terminal-Bench 4.0.0Agentic20.3% ±1.616 / 23
Terminal-Bench 4.1.0Agentic19.7% ±1.616 / 23
cursorBench40Coding41.4%15 / 21
ApprenticeBench: end-to-end computer use, continual learning, and long-horizon agency on a real accounts-payable jobAgentic13.0%13 / 18
FrontierSWE v2Coding25.3%16 / 22
Vibe Code Bench 1-100Coding14.8% ±3.219 / 26
Artificial Analysis AnalystAgentAgentic41.3%21 / 26
LegalBenchLegal86.3% ±0.414 / 149
Legal Research BenchLegal48.1% ±3.512 / 90
CorpFin v2Finance66.2% ±0.920 / 131
MedScribeHealthcare86.5% ±2.022 / 109
Tax Agent BenchFinance33.7% ±3.418 / 82
ARC-AGI-1 (public eval)Reasoning96.3%26 / 116
ARC-AGI-2 (public eval)Reasoning68.3%28 / 122
ProgramBenchCoding1.0% ±0.716 / 65
Public Benefits Bench v1.1Public benefits66.8% ±1.217 / 50
Terminal-Bench-ScienceScience11.4% ±3.815 / 44
BioMysteryBenchScience72.2%11 / 30
MedCodeHealthcare44.7% ±2.340 / 106
Finance Agent (v2)Finance53.7% ±0.732 / 79
EMBFinance62.7% ±3.131 / 75
MysteryMechanismScience30.6% ±3.114 / 30
MortgageTaxFinance64.2% ±1.047 / 97
TaxEval v2Finance71.1% ±0.983 / 143
CyberBench v1.1Other67.7% ±5.930 / 48
Time Horizon Index: KSPOther5.8%12 / 13
FrontierCode 1.1 ExtendedCoding61.3%—
Artificial Analysis EnterpriseOps-GymAgentic48.3%—

Sources

BenchLMCC BY-NC 4.0OpenRouterTerms restrict reuseEpoch AICC BYVals AINo licence statedLiveBenchNo licence statedARC Prize FoundationTerms restrict reuseTerminal-BenchNo licence statedLMArenaCC BY 4.0models.devMIT

History

5 releases
Index69.5 → 67.8
23 Oct 202612 Nov 2026
Output / 1M$6.00 → $6.00
23 Oct 202612 Nov 2026

Questions

How much does Grok 4.6 cost?
$2.00 per 1M input tokens, $6.00 per 1M output tokens.
How good is Grok 4.6?
Its index is 67.8 out of 100. That is place 30 of 283 ranked models.
What is Grok 4.6 best at?
Coding: 67.7 out of 100, place 28 of 209.
How fast is Grok 4.6?
56 output tokens per second, 2.8 s to the first token.
What is the context window of Grok 4.6?
500K tokens.
Is Grok 4.6 open source?
No. Its weights are not published.
When was Grok 4.6 released?
August 2026.

Badge

Grok 4.6 on Inference360: place 30 of 283, index 67.8
HTML
<a href="https://inference360.ai/models/grok-4-6"><img src="https://inference360.ai/badges/grok-4-6.svg" alt="Grok 4.6 on Inference360: place 30 of 283, index 67.8" height="20"></a>
Markdown
[![Grok 4.6 on Inference360: place 30 of 283, index 67.8](https://inference360.ai/badges/grok-4-6.svg)](https://inference360.ai/models/grok-4-6)

Compare with

Gemini 3.8 Flash67.8Claude Opus 4.867.7Ling 3.1 Flash67.2Step 5 Preview68.4Grok 4.769.1

Lower price

Ling 3.1 Flash67.2 · $0.135/1MDeepSeek V4.1 Flash 101768.5 · $0.40/1MClaude Haiku 5.566.3 · $0.50/1M

More from xAI

All models
Grok LatestUnrankedGrok 4.563.7Grok 4.769.1Grok Build 0.152.7Grok 4.353.3