AIEvals

Models

Grok 4.6

Specifications and API prices for Grok 4.6, recorded by Artificial Analysis.

xAIReleased 12 August 2026Context 500,000 tokensOutput Not published in the checked sourcePrivate weights$2.00 in / $6.00 out per Mtok

Family Grok 4.644 benchmarks with a published resultNo first place on the independent columns among the 46 models this site listsRead 17 October 2026

Where it stands

Rank compares tracked models with a score. Published shows the source’s rank and field size, when available.

BenchmarkScoreRankPublishedLeaderBehind
Vals IndexVals AI52.09%19of 3320of 41Gemini 4 Argon16.81 pts
Artificial Analysis Intelligence IndexArtificial Analysis4416of 42no list readClaude Opus 5.514 pts
Terminal-Bench 4.0Artificial Analysis17.17%26of 40no list readClaude Sonnet 5.546.47 pts
SWE-bench (Verified)Vals AI95.60%4of 234of 88Claude Opus 51.40 pts
LiveCodeBenchVals AI88.22%7of 217of 143Claude Fable 5.12.30 pts
GPQA DiamondVals AI94.70%3of 233of 138Gemini 3.1 Pro0.75 pts
MMLU ProVals AI89.40%9of 2411of 138Claude Fable 5.12.98 pts
Vibe Code Bench v1.1Vals AI76.24%26of 3828of 106Claude Sonnet 5.516.15 pts
Terminal-Bench Science 0.1Artificial Analysis6.19%17of 24no list readGPT-6 Astra57.14 pts
τ³-BankingArtificial Analysis43.30%12of 30no list readQwen3.8 Max8.04 pts
AA-AnalystAgentArtificial Analysis41.25%12of 18no list readGemini 3.7 Flash18.75 pts
MLCR-AAArtificial Analysis12.22%34of 37no list readClaude Sonnet 5.562.78 pts
ARC-AGI-3ARC Prize · Standard2.11%6of 11no list readGPT-6 Astra60.60 pts
Vals RSI IndexVals AI25.46%7of 207of 23Claude Opus 5.511.85 pts
Vibe Code Bench 1–100Vals AI14.75%15of 2015of 20Claude Opus 5.515.61 pts
EnterpriseOps-Gym-AAArtificial Analysis48.34%4of 22no list readClaude Fable 52.78 pts
AutomationBench-AAArtificial Analysis67.00%6of 40no list readGemini 4 Argon10.51 pts
SciCodeArtificial Analysis53.01%31of 40no list readClaude Opus 5.513.89 pts
GDP.pdfArtificial Analysis17.20%28of 40no list readGPT-6.1 Sol13.80 pts
CritPtArtificial Analysis19.71%22of 42no list readGPT-5.6 Sol12.58 pts
AA-OmniscienceArtificial Analysis29.3212of 42no list readClaude Opus 5.517.1
AA-LCR v1.1Artificial Analysis81.00%24of 42no list readKimi K37.67 pts
Terminal-Bench 2.1Vals AI78.28%13of 3513of 75Claude Opus 5.59.36 pts
ProofBench v1.1Vals AI51.00%23of 3427of 46Claude Sonnet 5.549.00 pts
LegalBenchVals AI86.31%10of 2713of 149Claude Fable 52.25 pts
Harvey's Legal Agent BenchmarkVals AI15.83%4of 376of 73Muse Spark 1.29.59 pts
SkillsBenchVals AI55.77%11of 2013of 35DeepSeek V4.1 Flash14.03 pts
MysteryMechanismVals AI30.63%11of 2111of 21GPT-6 Astra22.52 pts
BioMysteryBenchVals AI72.22%8of 198of 21Claude Sonnet 5.58.89 pts
Code MigrationVals AI44.57%18of 3718of 71Claude Sonnet 5.525.26 pts
CyberBench v1.1Vals AI67.68%19of 3125of 43GPT-6 Sol10.30 pts
EMBVals AI62.73%25of 3728of 69Claude Fable 5.113.94 pts
Finance Agent (v2)Vals AI53.68%22of 3727of 73Gemini 4 Argon11.72 pts
IOIVals AI47.61%25of 3126of 38Gemini 4 Argon52.39 pts
Legal Research BenchVals AI48.08%8of 378of 72Muse Spark 1.3 Max7.21 pts
MedCodeVals AI44.71%22of 3538of 104Claude Opus 518.86 pts
MedScribeVals AI86.53%17of 3620of 106Claude Opus 5.54.90 pts
Public Benefits Bench v1.1Vals AI66.85%15of 2915of 45Claude Opus 510.08 pts
SAGEVals AI28.90%32of 3283of 90Claude Opus 4.727.20 pts
Tax Agent BenchVals AI70.79%7of 378of 64Claude Fable 5.16.85 pts
DeepSWE v1.1Datacurve67.48%9of 209of 28GPT-6 Astra6.64 pts
GDPval-AA v2.1Artificial Analysis1,632.2211of 42no list readClaude Opus 5.5213.95
AA-Briefcase v1.1Artificial Analysis1,549.2712of 41no list readClaude Opus 5.5272.58
Humanity's Last ExamArtificial Analysis44.07%22of 42no list readClaude Opus 5.517.28 pts

Nearby on Vals Index

The two models above and below on the Vals Index (Vals AI), on every column where this model has a published number.

ModelVals Index%Artificial Analysis Intelligence IndexindexTerminal-Bench 4.0%SWE-bench%LiveCodeBench%GPQA Diamond%MMLU Pro%Vibe Code Bench v1.1%Terminal-Bench Science 0.1%
MiMo-V2.6-FlashXiaomi53.23%3822.73%––––78.96%–
GPT-5.6 TerraOpenAI53.09%4235.35%95.40%85.93%90.91%86.66%74.59%13.81%
Grok 4.6xAI52.09%4417.17%95.60%88.22%94.70%89.40%76.24%6.19%
Claude Sonnet 5Anthropic51.77%3814.14%79.60%82.43%88.89%87.55%81.33%–
GPT-5.6 LunaOpenAI51.69%3711.62%93.00%–91.67%86.04%77.06%3.33%
Bold marks a figure better than this model’s on the same column, whichever model it belongs to. Ranks cover the models shown from the stated source.

Cost and speed, as published

Additional measurements from Artificial Analysis’ own row. Read 30 September 2026.

As publishedCost per taskOutput speed
Grok 4.6 (xhigh)$2.3267.6 tok/s

Results by benchmark

Independent and vendor results retain their own harnesses and effort settings.

BenchmarkIndependent runAs reported by a publisher
τ³-BankingTool use43.30%Artificial Analysisnot published
AA-AnalystAgentData analysis41.25%Artificial Analysisnot published
MLCR-AAMedical reasoning12.22%Artificial Analysisnot published
ARC-AGI-3Interactive reasoning2.11%ARC Prize · Standardnot published
Vals RSI IndexAutonomous research25.46%Vals AInot published
Vibe Code Bench 1–100Application development14.75%Vals AInot published
EnterpriseOps-Gym-AAEnterprise workflows48.34%Artificial Analysisnot published
AutomationBench-AAWorkflow automation67.00%Artificial Analysisnot published
SciCodeScientific coding53.01%Artificial Analysisnot published
GDP.pdfDocument reasoning17.20%Artificial Analysisnot published
CritPtScientific reasoning19.71%Artificial Analysisnot published
AA-OmniscienceFactual knowledge29.32Artificial Analysisnot published
AA-LCR v1.1Long-context reasoning81.00%Artificial Analysisnot published
Terminal-Bench 4.0Agentic coding17.17%Artificial Analysis
17.17%Vals AI
not published
Terminal-Bench 2.1Agentic coding78.28%Vals AI
88.01%Artificial Analysis
not published
SWE-bench (Verified)Software engineering95.60%Vals AInot published
LiveCodeBenchCode generation88.22%Vals AInot published
GPQA DiamondScience reasoning94.70%Vals AI
93.54%Artificial Analysis
not published
MMLU ProKnowledge89.40%Vals AInot published
Vibe Code Bench v1.1Product coding76.24%Vals AInot published
Terminal-Bench Science 0.1Agentic science6.19%Artificial Analysis
11.43%Vals AI
not published
ProofBench v1.1Mathematics51.00%Vals AInot published
LegalBenchProfessional knowledge86.31%Vals AInot published
Harvey's Legal Agent BenchmarkProfessional knowledge15.83%Vals AInot published
SkillsBenchAgentic tooling55.77%Vals AInot published
MysteryMechanismScientific reasoning30.63%Vals AInot published
Vals IndexComposite52.09%Vals AInot published
BioMysteryBenchAgentic science72.22%Vals AInot published
Code MigrationAgentic coding44.57%Vals AInot published
CyberBench v1.1Security67.68%Vals AInot published
EMBFinancial modelling62.73%Vals AInot published
Finance Agent (v2)Financial analysis53.68%Vals AInot published
IOICompetitive programming47.61%Vals AInot published
Legal Research BenchProfessional knowledge48.08%Vals AInot published
MedCodeHealthcare administration44.71%Vals AInot published
MedScribeHealthcare administration86.53%Vals AInot published
Public Benefits Bench v1.1Public services66.85%Vals AInot published
SAGEEducation28.90%Vals AInot published
Tax Agent BenchProfessional knowledge70.79%Vals AInot published
DeepSWE v1.1Software engineering67.48%Datacurvenot published
GDPval-AA v2.1Knowledge work1,632.22Artificial Analysisnot published
AA-Briefcase v1.1Knowledge work1,549.27Artificial Analysisnot published
Humanity's Last ExamMultidisciplinary reasoning44.07%Artificial Analysisnot published
Artificial Analysis Intelligence IndexComposite44Artificial Analysisnot published
Independent column: the suite named under each figure, one harness per suite. Publisher column: each lab’s own table, reproduced on the benchmark’s page. Publisher ranks compare reported scores in the same table row only; ties share a rank and unreported cells are excluded.

Sources

Read on 17 October 2026.