RankingEffort benchmarks

Grok 4

Compare models

15 measured results · 2 effort settings · 14 benchmark/harness combinations

Change model

Compare with

Choose a second model and its effort to see both profiles and benchmark differences below.

Compare with

Capability profile

One configuration, seven capabilities

Comparison score · 0–100
Grok 4 · Reasoning capability radarComparison scores from 0 to 100. Unknown capabilities have no point; lines stop at gaps. Hollow points are preliminary. Exact values and support labels follow the chart.AgenticHard reasoningCodingHuman prefKnowledgeMultimodalLong context50100Grok 4 · Reasoning · Agentic: 13.7 · PreliminaryGrok 4 · Reasoning · Hard reasoning: 18.7 · SupportedGrok 4 · Reasoning · Coding: 23.7 · SupportedGrok 4 · Reasoning · Knowledge: 23.7 · SupportedGrok 4 · Reasoning · Multimodal: 11.0 · PreliminaryGrok 4 · Reasoning · Long context: 22.3 · Preliminary

Gaps are unknown, not zero. Hollow points are preliminary.

  • Agentic

    A13.7Preliminary
  • Hard reasoning

    A18.7Supported
  • Coding

    A23.7Supported
  • Human pref

    Unknown
  • Knowledge

    A23.7Supported
  • Multimodal

    A11.0Preliminary
  • Long context

    A22.3Preliminary

Each profile belongs to the named configuration. Capability scores use a shared panel for all axes, distinct from the overall ranking panel; they are not accuracy percentages or direct head-to-head wins.

Capability score evidence

Grok 4 · Reasoning: 8.6 overall comparison score. Meets the overall evidence requirements.

6 areas measured. The available task mix affects this estimate; missing areas remain unknown.

Which sources carry the weight?
  • Artificial Analysis · 53.5% direct comparison weight · independent of the model provider
  • ARC Prize · 46.5% direct comparison weight · independent of the model provider
Which benchmark families carry the weight?
  • ARC-AGI · 46.5% direct comparison weight
  • MMLU-Pro · 4.2% direct comparison weight
  • LiveCodeBench · 4.2% direct comparison weight
  • Aime · 4.2% direct comparison weight
  • math500 · 4.2% direct comparison weight
  • Other families (9) · 36.6% direct comparison weight

These shares describe direct comparison weight, not fractions of the final score or confidence. Source balancing does not give every family equal weight or remove correlations between overlapping sources.

How sensitive is this score?

The audit removes one source at a time and refits against the same overall reference panel. These published results use the matching dataset audited 2026-11-22.

Both ranks use the same surviving configurations for each removal. The cohort starts with 602 configurations supported in the original audit, including separate effort settings. These are diagnostic positions, not the default leaderboard ranks. Eligibility is not rechecked after removal.

What changes when a source is removed?
Removed sourceRank before → afterRank movementCohort remaining / lost supportScore after (change)
Artificial Analysis183 → 182Up 1230 remaining / 372 lost6.6 (-2.0 points)
ARC Prize278 → 258Up 20602 remaining / 0 lost13.1 (+4.4 points)
Datacurve278 → 279Down 1602 remaining / 0 lost6.9 (-1.7 points)
LiveCodeBench278 → 277Up 1602 remaining / 0 lost8.6 (-0.0 points)
Scale278 → 278No movement602 remaining / 0 lost9.6 (+1.0 points)
Terminal-Bench278 → 278No movement602 remaining / 0 lost8.6 (-0.1 points)
Cognition278 → 276Up 2602 remaining / 0 lost8.2 (-0.4 points)
LiveBench277 → 277No movement601 remaining / 1 lost9.0 (+0.4 points)
SWE-rebench278 → 280Down 2602 remaining / 0 lost8.7 (+0.1 points)
Epoch AI271 → 275Down 4588 remaining / 14 lost8.7 (+0.1 points)
Vals AI278 → 279Down 1601 remaining / 1 lost8.2 (-0.4 points)
Provider reports · Anthropic278 → 278No movement602 remaining / 0 lost8.6 (-0.0 points)
Provider reports · Moonshot AI278 → 278No movement602 remaining / 0 lost8.4 (-0.2 points)
Provider reports · Meta278 → 278No movement602 remaining / 0 lost8.6 (-0.0 points)
Provider reports · Mistral278 → 278No movement602 remaining / 0 lost8.6 (+0.0 points)
Provider reports · OpenAI278 → 278No movement602 remaining / 0 lost8.7 (+0.1 points)
All provider reports278 → 278No movement602 remaining / 0 lost8.5 (-0.2 points)

Lost support means a configuration no longer has a comparison path to the complete reference panel. Its rank is unavailable. A surviving path alone does not establish leaderboard eligibility.

Overlapping sources may remain correlated even with source balancing. Leave-one-source-out movement is a sensitivity diagnostic, not a spread of source-specific ranks or a confidence measure. Scoring methodology · Dated configuration audit

Every benchmark family
Removed evidenceRecomputed scoreChange
arc-agi12.6+4.0
deepswe6.9-1.7
mmmu-pro9.5+0.9
epoch-game-puzzles8.2-0.4
frontiercode8.2-0.4
simpleqa9.0+0.3
arc-agi-38.3-0.3
gpqa8.3-0.3
hle8.9+0.3
gmmlu8.4-0.2
ifbench8.8+0.2
tau-bench8.8+0.2
livebench-instructions8.8+0.2
aime8.4-0.2
enigma-eval8.8+0.2
mmlu-pro8.8+0.2
multi-swe-bench8.5-0.2
terminal-bench8.5-0.1
livebench-language8.7+0.1
apex-agents8.7+0.1
livebench-coding8.5-0.1
swe-atlas-qna8.5-0.1
aa-lcr8.5-0.1
vibe-code8.5-0.1
critpt8.5-0.1
enterprise-ops8.5-0.1
automationbench8.5-0.1
gdp-pdf8.7+0.1
swe-rebench8.7+0.1
vals-code-migration8.6-0.1
gdpval8.7+0.1
terminal-bench-science8.7+0.1
math5008.7+0.1
livebench-reasoning8.6-0.1
briefcase8.6-0.0
vals-finance-agent8.6-0.0
itbench8.6-0.0
vals-legal-research8.6-0.0
vals-tax-agent8.6-0.0
analyst-agent8.7+0.0
swe-bench-pro8.7+0.0
swe-atlas-test-writing8.6-0.0
vals-excel-modeling8.6-0.0
livecodebench8.6-0.0
omniscience8.6-0.0
frontiermath8.6+0.0
swe-atlas-refactoring8.6-0.0
livebench-math8.6+0.0
livebench-data8.6+0.0
harvey8.6+0.0
What evidence is missing from the fit?

Across all collected settings for this model: 15 observations, 15 contributing, 0 matched without graph weight and 0 excluded. Counts do not establish rank eligibility.

    Inspect every observation and exclusion →

    Benchmark results

    These are published benchmark scores. Supported effort estimates also appear on the Capability leaderboard. Results from different harnesses occupy different rows. Missing cells stay unknown; no scores are copied between effort levels.

    Reviewed retained observations are in use. October 2026 Real-SWE board has no 95% confidence whisker, so these scores stay at the 2026-10-18 retrieval. Original source retrieval 2026-10-18; source updated 2026-10-18.

    Benchmark / sourceThinkingReasoning
    AIME 2025Artificial Analysis · Artificial Analysis published evaluation; see source for benchmark-specific settings
    Benchmark details

    Independently benchmarked by Artificial Analysis · Evaluation results measured independently by Artificial Analysis

    Unit: percent. Source collected 2026-11-22.

    Not reported
    92.67%
    Reported settings

    Grok 4; Reported effort: reasoning; Model source: https://artificialanalysis.ai/models/grok-4; Mode from evaluator isReasoning metadata; effort budget not specified

    AA-LCR v1.1Artificial Analysis · Artificial Analysis published evaluation; see source for benchmark-specific settings
    Benchmark details

    Independently benchmarked by Artificial Analysis · Evaluation results measured independently by Artificial Analysis

    Unit: percent. Source collected 2026-11-22.

    Not reported
    68%
    Reported settings

    Grok 4; Reported effort: reasoning; Model source: https://artificialanalysis.ai/models/grok-4; Mode from evaluator isReasoning metadata; effort budget not specified

    CritPtArtificial Analysis · Artificial Analysis published evaluation; see source for benchmark-specific settings
    Benchmark details

    Benchmark developed by Argonne and UIUC, with contributions from 60+ researchers globally · Independently benchmarked by Artificial Analysis · Evaluation results measured independently by Artificial Analysis

    Unit: percent. Source collected 2026-11-22.

    Not reported
    2%
    Reported settings

    Grok 4; Reported effort: reasoning; Model source: https://artificialanalysis.ai/models/grok-4; Mode from evaluator isReasoning metadata; effort budget not specified

    Global-MMLU-LiteArtificial Analysis · Artificial Analysis published evaluation; see source for benchmark-specific settings
    Benchmark details

    Independently benchmarked by Artificial Analysis · Evaluation results measured independently by Artificial Analysis

    Unit: percent. Source collected 2026-11-22.

    Not reported
    89.49%
    Reported settings

    Grok 4; Reported effort: reasoning; Model source: https://artificialanalysis.ai/models/grok-4; Mode from evaluator isReasoning metadata; effort budget not specified

    GPQA DiamondArtificial Analysis · Artificial Analysis published evaluation; see source for benchmark-specific settings
    Benchmark details

    Independently benchmarked by Artificial Analysis · Evaluation results measured independently by Artificial Analysis

    Unit: percent. Source collected 2026-11-22.

    Not reported
    87.68%
    Reported settings

    Grok 4; Reported effort: reasoning; Model source: https://artificialanalysis.ai/models/grok-4; Mode from evaluator isReasoning metadata; effort budget not specified

    Humanity's Last ExamArtificial Analysis · Artificial Analysis published evaluation; see source for benchmark-specific settings
    Benchmark details

    Independently benchmarked by Artificial Analysis · Evaluation results measured independently by Artificial Analysis

    Unit: percent. Source collected 2026-11-22.

    Not reported
    26.69%
    Reported settings

    Grok 4; Reported effort: reasoning; Model source: https://artificialanalysis.ai/models/grok-4; Mode from evaluator isReasoning metadata; effort budget not specified

    IFBenchArtificial Analysis · Artificial Analysis published evaluation; see source for benchmark-specific settings
    Benchmark details

    Independently benchmarked by Artificial Analysis · Evaluation results measured independently by Artificial Analysis

    Unit: percent. Source collected 2026-11-22.

    Not reported
    53.67%
    Reported settings

    Grok 4; Reported effort: reasoning; Model source: https://artificialanalysis.ai/models/grok-4; Mode from evaluator isReasoning metadata; effort budget not specified

    LiveCodeBenchArtificial Analysis · Artificial Analysis published evaluation; see source for benchmark-specific settings
    Benchmark details

    Independently benchmarked by Artificial Analysis · Evaluation results measured independently by Artificial Analysis

    Unit: percent. Source collected 2026-11-22.

    Not reported
    81.9%
    Reported settings

    Grok 4; Reported effort: reasoning; Model source: https://artificialanalysis.ai/models/grok-4; Mode from evaluator isReasoning metadata; effort budget not specified

    MATH-500Artificial Analysis · Artificial Analysis published evaluation; see source for benchmark-specific settings
    Benchmark details

    Independently benchmarked by Artificial Analysis · Evaluation results measured independently by Artificial Analysis

    Unit: percent. Source collected 2026-11-22.

    Not reported
    99%
    Reported settings

    Grok 4; Reported effort: reasoning; Model source: https://artificialanalysis.ai/models/grok-4; Mode from evaluator isReasoning metadata; effort budget not specified

    MMLU-ProArtificial Analysis · Artificial Analysis published evaluation; see source for benchmark-specific settings
    Benchmark details

    Independently benchmarked by Artificial Analysis · Evaluation results measured independently by Artificial Analysis

    Unit: percent. Source collected 2026-11-22.

    Not reported
    86.58%
    Reported settings

    Grok 4; Reported effort: reasoning; Model source: https://artificialanalysis.ai/models/grok-4; Mode from evaluator isReasoning metadata; effort budget not specified

    MMMU-ProArtificial Analysis · Artificial Analysis published evaluation; see source for benchmark-specific settings
    Benchmark details

    Independently benchmarked by Artificial Analysis · Evaluation results measured independently by Artificial Analysis

    Unit: percent. Source collected 2026-11-22.

    Not reported
    68.84%
    Reported settings

    Grok 4; Reported effort: reasoning; Model source: https://artificialanalysis.ai/models/grok-4; Mode from evaluator isReasoning metadata; effort budget not specified

    𝜏²-Bench TelecomArtificial Analysis · Artificial Analysis published evaluation; see source for benchmark-specific settings
    Benchmark details

    Independently benchmarked by Artificial Analysis · Evaluation results measured independently by Artificial Analysis

    Unit: percent. Source collected 2026-11-22.

    Not reported
    74.85%
    Reported settings

    Grok 4; Reported effort: reasoning; Model source: https://artificialanalysis.ai/models/grok-4; Mode from evaluator isReasoning metadata; effort budget not specified

    Terminal-Bench HardArtificial Analysis · Artificial Analysis published evaluation; see source for benchmark-specific settings
    Benchmark details

    Independently benchmarked by Artificial Analysis · Evaluation results measured independently by Artificial Analysis

    Unit: percent. Source collected 2026-11-22.

    Not reported
    37.88%
    Reported settings

    Grok 4; Reported effort: reasoning; Model source: https://artificialanalysis.ai/models/grok-4; Mode from evaluator isReasoning metadata; effort budget not specified

    ARC-AGI-2ARC Prize · v2 Semi-Private
    Benchmark details

    Original public board. Configuration and task coverage may differ; inspect source.

    Unit: percent. Source collected 2026-11-22.

    15.98%
    Reported settings

    Grok 4 (Thinking); Source model ID: grok-4-0709; 120 tasks

    5.28%
    Reported settings

    Grok 4 (Fast Reasoning); Source model ID: grok-4-fast-reasoning; 120 tasks

    What these comparisons mean

    Each source reports its own effort labels. Matching labels do not establish equal compute budgets or identical fallback behavior. Open Reported settings to inspect the exact model name, agent and task coverage. Multiple results at one level remain visible rather than selecting the best. Source-reported None or Non-reasoning labels are not treated as evidence that an API supports those settings.

    These measurements use their own dated source collection. The leaderboard fits identified configurations jointly; it contains no pooled model ratings. Explicit thinking budgets are separate configurations. Read the methodology.

    Download all effort benchmark results · 10,184 results across 381 models · Collected 2026-11-22