Compare · ModelsLive · 3 picked · head to head
Grok 3 Mini Beta vs Grok 4 vs GPT-5.1
Side by side · benchmarks, pricing, and signals you can act on.
Winner summary
GPT-5.1 wins on 13/19 benchmarks
GPT-5.1 wins 13 of 19 shared benchmarks. Leads in knowledge · math · reasoning.
Category leads
knowledge·GPT-5.1language·Grok 3 Mini Betamath·GPT-5.1reasoning·GPT-5.1coding·GPT-5.1agentic·GPT-5.1arena·GPT-5.1
Hype vs Reality
Attention vs performance
Grok 3 Mini Beta
#30 by perf·no signal
Grok 4
#73 by perf·no signal
GPT-5.1
#97 by perf·no signal
Best value
Grok 3 Mini Beta
18.4x better value than GPT-5.1
Grok 3 Mini Beta
162.0 pts/$
$0.40/M
Grok 4
6.1 pts/$
$9.00/M
GPT-5.1
8.8 pts/$
$5.63/M
Vendor risk
Who is behind the model
xAI
$250.0B·Tier 1
xAI
$250.0B·Tier 1
OpenAI
$840.0B·Tier 1
Head to head
19 benchmarks · 3 models
Grok 3 Mini BetaGrok 4GPT-5.1
HELM · GPQA
Grok 4 leads by +5.1
Grok 3 Mini Beta
67.5
Grok 4
72.6
GPT-5.1
44.2
HELM · IFEval
Grok 3 Mini Beta leads by +0.2
Grok 3 Mini Beta
95.1
Grok 4
94.9
GPT-5.1
93.5
HELM · MMLU-Pro
Grok 4 leads by +5.2
Grok 3 Mini Beta
79.9
Grok 4
85.1
GPT-5.1
57.9
HELM · Omni-MATH
Grok 4 leads by +13.9
Grok 3 Mini Beta
31.8
Grok 4
60.3
GPT-5.1
46.4
HELM · WildBench
GPT-5.1 leads by +6.6
Grok 3 Mini Beta
65.1
Grok 4
79.7
GPT-5.1
86.3
Aider polyglot
Grok 4 leads by +30.3
Aider Polyglot · measures how well AI models can edit code across multiple programming languages using the Aider coding assistant framework.
Grok 3 Mini Beta
49.3
Grok 4
79.6
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.
Grok 4
15.2
GPT-5.1
17.5
ARC-AGI
GPT-5.1 leads by +6.1
ARC-AGI · the original Abstraction and Reasoning Corpus, testing whether AI can solve novel visual pattern recognition tasks without memorization.
Grok 4
66.7
GPT-5.1
72.8
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.
Grok 4
16.0
GPT-5.1
17.6
Chatbot Arena Elo · Overall
GPT-5.1 leads by +81.1
Grok 3 Mini Beta
1357.4
GPT-5.1
1438.5
Chess Puzzles
GPT-5.1 leads by +4.0
Chess Puzzles · tests strategic and tactical reasoning by having models solve chess puzzle positions, evaluating lookahead and pattern recognition abilities.
Grok 4
28.0
GPT-5.1
32.0
FrontierMath-2025-02-28-Private
GPT-5.1 leads by +11.4
FrontierMath (Feb 2025) · original research-level math problems created by mathematicians, testing capabilities at the boundary of current AI mathematical reasoning.
Grok 4
19.7
GPT-5.1
31.0
FrontierMath-Tier-4-2025-07-01-Private
GPT-5.1 leads by +10.4
FrontierMath Tier 4 (Jul 2025) · the most challenging tier of frontier mathematics, containing problems that push the absolute limits of AI mathematical reasoning.
Grok 4
2.1
GPT-5.1
12.5
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.
Grok 4
82.7
GPT-5.1
83.5
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.
Grok 4
84.0
GPT-5.1
88.6
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.
Grok 4
52.6
GPT-5.1
43.8
SimpleQA Verified
GPT-5.1 leads by +1.0
SimpleQA Verified · short factual questions with verified answers, measuring factual accuracy and the tendency to hallucinate or provide incorrect information.
Grok 4
47.9
GPT-5.1
48.9
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.
Grok 4
27.2
GPT-5.1
47.6
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.
Grok 4
45.7
GPT-5.1
60.8
Full benchmark table
| Benchmark | Grok 3 Mini Beta | Grok 4 | GPT-5.1 |
|---|---|---|---|
HELM · GPQA | 67.5 | 72.6 | 44.2 |
HELM · IFEval | 95.1 | 94.9 | 93.5 |
HELM · MMLU-Pro | 79.9 | 85.1 | 57.9 |
HELM · Omni-MATH | 31.8 | 60.3 | 46.4 |
HELM · WildBench | 65.1 | 79.7 | 86.3 |
Aider polyglot Aider Polyglot · measures how well AI models can edit code across multiple programming languages using the Aider coding assistant framework. | 49.3 | 79.6 | — |
APEX-Agents APEX-Agents · evaluates AI agents on complex, multi-step tasks requiring planning, tool use, and autonomous decision-making in realistic environments. | — | 15.2 | 17.5 |
ARC-AGI ARC-AGI · the original Abstraction and Reasoning Corpus, testing whether AI can solve novel visual pattern recognition tasks without memorization. | — | 66.7 | 72.8 |
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. | — | 16.0 | 17.6 |
Chatbot Arena Elo · Overall | 1357.4 | — | 1438.5 |
Chess Puzzles Chess Puzzles · tests strategic and tactical reasoning by having models solve chess puzzle positions, evaluating lookahead and pattern recognition abilities. | — | 28.0 | 32.0 |
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. | — | 19.7 | 31.0 |
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. | — | 2.1 | 12.5 |
GPQA diamond Graduate-Level Google-Proof QA (Diamond set) · expert-crafted questions in physics, biology, and chemistry that are difficult even for domain PhDs. | — | 82.7 | 83.5 |
OTIS Mock AIME 2024-2025 OTIS Mock AIME 2024-2025 · simulated American Invitational Mathematics Examination problems testing advanced problem-solving skills. | — | 84.0 | 88.6 |
SimpleBench SimpleBench · tests fundamental reasoning capabilities with straightforward problems designed to expose gaps in basic logical and spatial thinking. | — | 52.6 | 43.8 |
SimpleQA Verified SimpleQA Verified · short factual questions with verified answers, measuring factual accuracy and the tendency to hallucinate or provide incorrect information. | — | 47.9 | 48.9 |
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. | — | 27.2 | 47.6 |
WeirdML WeirdML · tests models on unusual and adversarial machine learning tasks that require creative problem-solving beyond standard patterns. | — | 45.7 | 60.8 |
Pricing · per 1M tokens · projected $/mo at 10M tokens
| Model | Input | Output | Context | Projected $/mo |
|---|---|---|---|---|
| $0.30 | $0.50 | 131K tokens (~66 books) | $3.50 | |
| $3.00 | $15.00 | 256K tokens (~128 books) | $60.00 | |
| $1.25 | $10.00 | 400K tokens (~200 books) | $34.38 |