AIEvals

Models

Grok 4.7

The Grok 4.7 release, with published reasoning, coding and agentic evaluations.

xAIReleased 21 September 2026Context 500k tokensOutput 500k tokensPrivate weights$2.00 in / $6.00 out per Mtok

Family Grok 4.735 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 AI54.95%14of 3314of 41Gemini 4 Argon13.95 pts
Artificial Analysis Intelligence IndexArtificial Analysis4612of 42no list readClaude Opus 5.512 pts
Terminal-Bench 4.0Artificial Analysis25.76%20of 40no list readClaude Sonnet 5.537.88 pts
Vibe Code Bench v1.1Vals AI86.17%10of 3810of 106Claude Sonnet 5.56.22 pts
ProgramBenchVals AI0.50%13of 2913of 54Claude Opus 5.518.00 pts
ITBench-AAArtificial Analysis42.08%13of 18no list readGPT-5.6 Sol14.13 pts
MLCR-AAArtificial Analysis15.00%31of 37no list readClaude Sonnet 5.560.00 pts
Vals RSI IndexVals AI20.20%14of 2014of 23Claude Opus 5.517.11 pts
AutomationBench-AAArtificial Analysis65.56%7of 40no list readGemini 4 Argon11.95 pts
SciCodeArtificial Analysis57.41%13of 40no list readClaude Opus 5.59.49 pts
GDP.pdfArtificial Analysis20.00%23of 40no list readGPT-6.1 Sol11.00 pts
CritPtArtificial Analysis17.71%28of 42no list readGPT-5.6 Sol14.58 pts
AA-OmniscienceArtificial Analysis32.039of 42no list readClaude Opus 5.514.39
AA-LCR v1.1Artificial Analysis76.67%40of 42no list readKimi K312.00 pts
Terminal-Bench 2.1Vals AI73.41%20of 3522of 75Claude Opus 5.514.23 pts
ProofBench v1.1Vals AI26.00%29of 3434of 46Claude Sonnet 5.574.00 pts
LegalBenchVals AI84.39%18of 2731of 149Claude Fable 54.17 pts
Harvey's Legal Agent BenchmarkVals AI12.50%6of 378of 73Muse Spark 1.212.92 pts
MysteryMechanismVals AI25.23%13of 2113of 21GPT-6 Astra27.92 pts
BioMysteryBenchVals AI69.26%11of 1911of 21Claude Sonnet 5.511.85 pts
Code MigrationVals AI44.82%17of 3717of 71Claude Sonnet 5.525.01 pts
CyberBench v1.1Vals AI69.46%17of 3121of 43GPT-6 Sol8.52 pts
EMBVals AI66.99%17of 3717of 69Claude Fable 5.19.68 pts
Finance Agent (v2)Vals AI52.25%25of 3730of 73Gemini 4 Argon13.15 pts
IOIVals AI57.72%16of 3116of 38Gemini 4 Argon42.28 pts
Legal Research BenchVals AI47.12%12of 3712of 72Muse Spark 1.3 Max8.17 pts
MedCodeVals AI49.55%11of 3520of 104Claude Opus 514.02 pts
MedScribeVals AI89.38%6of 366of 106Claude Opus 5.52.05 pts
Public Benefits Bench v1.1Vals AI65.63%18of 2918of 45Claude Opus 511.30 pts
SAGEVals AI40.79%27of 3258of 90Claude Opus 4.715.31 pts
Tax Agent BenchVals AI65.60%14of 3716of 64Claude Fable 5.112.04 pts
GDPval-AA v2.1Artificial Analysis1,695.215of 42no list readClaude Opus 5.5150.96
AA-Briefcase v1.1Artificial Analysis1,657.25of 41no list readClaude Opus 5.5164.65
Humanity's Last ExamArtificial Analysis43.14%24of 42no list readClaude Opus 5.518.21 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%Vibe Code Bench v1.1%ProgramBench%
MiMo V2.6 ProXiaomi55.20%4634.85%85.22%0.50%
Claude Opus 4.8Anthropic55.10%4221.72%82.72%1.00%
Grok 4.7xAI54.95%4625.76%86.17%0.50%
Gemini 3.8 FlashGoogle54.83%4119.70%78.65%1.00%
GLM 5.3Z.ai53.51%4541.92%78.13%1.50%
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 4.7 (xhigh)$3.7473.6 tok/s

Results by benchmark

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

BenchmarkIndependent runAs reported by a publisher
ITBench-AAIT operations42.08%Artificial Analysisnot published
MLCR-AAMedical reasoning15.00%Artificial Analysisnot published
Vals RSI IndexAutonomous research20.20%Vals AInot published
AutomationBench-AAWorkflow automation65.56%Artificial Analysisnot published
SciCodeScientific coding57.41%Artificial Analysisnot published
GDP.pdfDocument reasoning20.00%Artificial Analysisnot published
CritPtScientific reasoning17.71%Artificial Analysisnot published
AA-OmniscienceFactual knowledge32.03Artificial Analysisnot published
AA-LCR v1.1Long-context reasoning76.67%Artificial Analysisnot published
Terminal-Bench 4.0Agentic coding25.76%Artificial Analysis
28.79%Vals AI
not published
Terminal-Bench 2.1Agentic coding73.41%Vals AInot published
Vibe Code Bench v1.1Product coding86.17%Vals AInot published
Terminal-Bench Science 0.1Agentic science
10.00%Vals AI
not published
ProgramBenchAgentic coding0.50%Vals AInot published
ProofBench v1.1Mathematics26.00%Vals AInot published
LegalBenchProfessional knowledge84.39%Vals AInot published
Harvey's Legal Agent BenchmarkProfessional knowledge12.50%Vals AInot published
MysteryMechanismScientific reasoning25.23%Vals AInot published
Vals IndexComposite54.95%Vals AInot published
BioMysteryBenchAgentic science69.26%Vals AInot published
Code MigrationAgentic coding44.82%Vals AInot published
CyberBench v1.1Security69.46%Vals AInot published
EMBFinancial modelling66.99%Vals AInot published
Finance Agent (v2)Financial analysis52.25%Vals AInot published
IOICompetitive programming57.72%Vals AInot published
Legal Research BenchProfessional knowledge47.12%Vals AInot published
MedCodeHealthcare administration49.55%Vals AInot published
MedScribeHealthcare administration89.38%Vals AInot published
Public Benefits Bench v1.1Public services65.63%Vals AInot published
SAGEEducation40.79%Vals AInot published
Tax Agent BenchProfessional knowledge65.60%Vals AInot published
GDPval-AA v2.1Knowledge work1,695.21Artificial Analysisnot published
AA-Briefcase v1.1Knowledge work1,657.2Artificial Analysisnot published
Humanity's Last ExamMultidisciplinary reasoning43.14%Artificial Analysisnot published
Artificial Analysis Intelligence IndexComposite46Artificial 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.