Compare · ModelsLive · 2 picked · head to head
GPT-5.1 vs Grok 4
Side by side · benchmarks, pricing, and signals you can act on.
Winner summary
GPT-5.1 wins on 12/18 benchmarks
GPT-5.1 wins 12 of 18 shared benchmarks. Leads in agentic · reasoning · knowledge.
Category leads
agentic·GPT-5.1reasoning·GPT-5.1knowledge·GPT-5.1math·GPT-5.1language·Grok 4coding·GPT-5.1
Hype vs Reality
Attention vs performance
GPT-5.1
#138 by perf·#4 by attention
Grok 4
#87 by perf·#15 by attention
Vendor risk
Who is behind the model
OpenAI
$840.0B·Tier 1
xAI
$250.0B·Tier 1
Head to head
18 benchmarks · 2 models
GPT-5.1Grok 4
APEX-Agents
GPT-5.1 leads by +2.3
APEX-Agents · evaluates AI agents on complex, multi-step tasks requiring planning, tool use, and autonomous decision-making in realistic environments.
GPT-5.1
17.5
Grok 4
15.2
ARC-AGI
GPT-5.1 leads by +6.2
ARC-AGI · the original Abstraction and Reasoning Corpus, testing whether AI can solve novel visual pattern recognition tasks without memorization.
GPT-5.1
72.8
Grok 4
66.7
ARC-AGI-2
GPT-5.1 leads by +1.7
ARC-AGI-2 · the second iteration of the Abstraction and Reasoning Corpus, testing novel pattern recognition and abstract reasoning without prior training data.
GPT-5.1
17.6
Grok 4
16.0
Chess Puzzles
GPT-5.1 leads by +4.2
Chess Puzzles · tests strategic and tactical reasoning by having models solve chess puzzle positions, evaluating lookahead and pattern recognition abilities.
GPT-5.1
28.4
Grok 4
24.2
DeepResearch Bench
Grok 4 leads by +4.5
DeepResearch Bench · evaluates AI on complex multi-step research tasks requiring information gathering, synthesis, and producing comprehensive analyses.
GPT-5.1
42.8
Grok 4
47.3
FrontierMath-2025-02-28-Private
GPT-5.1 leads by +20.0
FrontierMath (Feb 2025) · original research-level math problems created by mathematicians, testing capabilities at the boundary of current AI mathematical reasoning.
GPT-5.1
54.5
Grok 4
34.5
FrontierMath-Tier-4-2025-07-01-Private
GPT-5.1 leads by +17.4
FrontierMath Tier 4 (Jul 2025) · the most challenging tier of frontier mathematics, containing problems that push the absolute limits of AI mathematical reasoning.
GPT-5.1
20.8
Grok 4
3.5
GPQA diamond
GPT-5.1 leads by +0.8
Graduate-Level Google-Proof QA (Diamond set) · expert-crafted questions in physics, biology, and chemistry that are difficult even for domain PhDs.
GPT-5.1
83.5
Grok 4
82.7
HELM · GPQA
Grok 4 leads by +28.4
GPT-5.1
44.2
Grok 4
72.6
HELM · IFEval
Grok 4 leads by +1.4
GPT-5.1
93.5
Grok 4
94.9
HELM · MMLU-Pro
Grok 4 leads by +27.2
GPT-5.1
57.9
Grok 4
85.1
HELM · Omni-MATH
Grok 4 leads by +13.9
GPT-5.1
46.4
Grok 4
60.3
HELM · WildBench
GPT-5.1 leads by +6.6
GPT-5.1
86.3
Grok 4
79.7
OTIS Mock AIME 2024-2025
GPT-5.1 leads by +4.6
OTIS Mock AIME 2024-2025 · simulated American Invitational Mathematics Examination problems testing advanced problem-solving skills.
GPT-5.1
88.6
Grok 4
84.0
SimpleBench
Grok 4 leads by +8.8
SimpleBench · tests fundamental reasoning capabilities with straightforward problems designed to expose gaps in basic logical and spatial thinking.
GPT-5.1
43.8
Grok 4
52.6
SimpleQA Verified
GPT-5.1 leads by +0.1
SimpleQA Verified · short factual questions with verified answers, measuring factual accuracy and the tendency to hallucinate or provide incorrect information.
GPT-5.1
48.0
Grok 4
47.9
Terminal Bench
GPT-5.1 leads by +20.4
Terminal-Bench 2.0 · evaluates AI agents on real terminal-based coding tasks · writing scripts, debugging, running tests, and managing projects entirely through command-line interaction. Tests both code quality and terminal fluency. Claude Opus 4.7 scores 69.4%, demonstrating significant agentic terminal competence.
GPT-5.1
47.6
Grok 4
27.2
WeirdML
GPT-5.1 leads by +15.0
WeirdML · tests models on unusual and adversarial machine learning tasks that require creative problem-solving beyond standard patterns.
GPT-5.1
60.8
Grok 4
45.7
Full benchmark table
| Benchmark | GPT-5.1 | Grok 4 |
|---|---|---|
APEX-Agents APEX-Agents · evaluates AI agents on complex, multi-step tasks requiring planning, tool use, and autonomous decision-making in realistic environments. | 17.5 | 15.2 |
ARC-AGI ARC-AGI · the original Abstraction and Reasoning Corpus, testing whether AI can solve novel visual pattern recognition tasks without memorization. | 72.8 | 66.7 |
ARC-AGI-2 ARC-AGI-2 · the second iteration of the Abstraction and Reasoning Corpus, testing novel pattern recognition and abstract reasoning without prior training data. | 17.6 | 16.0 |
Chess Puzzles Chess Puzzles · tests strategic and tactical reasoning by having models solve chess puzzle positions, evaluating lookahead and pattern recognition abilities. | 28.4 | 24.2 |
DeepResearch Bench DeepResearch Bench · evaluates AI on complex multi-step research tasks requiring information gathering, synthesis, and producing comprehensive analyses. | 42.8 | 47.3 |
FrontierMath-2025-02-28-Private FrontierMath (Feb 2025) · original research-level math problems created by mathematicians, testing capabilities at the boundary of current AI mathematical reasoning. | 54.5 | 34.5 |
FrontierMath-Tier-4-2025-07-01-Private FrontierMath Tier 4 (Jul 2025) · the most challenging tier of frontier mathematics, containing problems that push the absolute limits of AI mathematical reasoning. | 20.8 | 3.5 |
GPQA diamond Graduate-Level Google-Proof QA (Diamond set) · expert-crafted questions in physics, biology, and chemistry that are difficult even for domain PhDs. | 83.5 | 82.7 |
HELM · GPQA | 44.2 | 72.6 |
HELM · IFEval | 93.5 | 94.9 |
HELM · MMLU-Pro | 57.9 | 85.1 |
HELM · Omni-MATH | 46.4 | 60.3 |
HELM · WildBench | 86.3 | 79.7 |
OTIS Mock AIME 2024-2025 OTIS Mock AIME 2024-2025 · simulated American Invitational Mathematics Examination problems testing advanced problem-solving skills. | 88.6 | 84.0 |
SimpleBench SimpleBench · tests fundamental reasoning capabilities with straightforward problems designed to expose gaps in basic logical and spatial thinking. | 43.8 | 52.6 |
SimpleQA Verified SimpleQA Verified · short factual questions with verified answers, measuring factual accuracy and the tendency to hallucinate or provide incorrect information. | 48.0 | 47.9 |
Terminal Bench Terminal-Bench 2.0 · evaluates AI agents on real terminal-based coding tasks · writing scripts, debugging, running tests, and managing projects entirely through command-line interaction. Tests both code quality and terminal fluency. Claude Opus 4.7 scores 69.4%, demonstrating significant agentic terminal competence. | 47.6 | 27.2 |
WeirdML WeirdML · tests models on unusual and adversarial machine learning tasks that require creative problem-solving beyond standard patterns. | 60.8 | 45.7 |
Pricing · per 1M tokens · projected $/mo at 10M tokens
People also compared