Compare · ModelsLive · 2 picked · head to head
Claude Haiku 4.5 vs GPT-5.4 Mini
Side by side · benchmarks, pricing, and signals you can act on.
Winner summary
GPT-5.4 Mini wins on 11/13 benchmarks
GPT-5.4 Mini wins 11 of 13 shared benchmarks. Leads in agentic · reasoning · knowledge.
Category leads
agentic·GPT-5.4 Minireasoning·GPT-5.4 Miniknowledge·GPT-5.4 Minigeneral·GPT-5.4 Minimath·GPT-5.4 Minicoding·GPT-5.4 Mini
Hype vs Reality
Attention vs performance
Claude Haiku 4.5
#231 by perf·#13 by attention
GPT-5.4 Mini
#209 by perf·#4 by attention
Best value
GPT-5.4 Mini
1.3x better value than Claude Haiku 4.5
Claude Haiku 4.5
11.3 pts/$
$3.00/M
GPT-5.4 Mini
14.3 pts/$
$2.63/M
Vendor risk
Who is behind the model
Anthropic
$965.0B·Tier 1
OpenAI
$840.0B·Tier 1
Head to head
13 benchmarks · 2 models
Claude Haiku 4.5GPT-5.4 Mini
APEX-Agents
GPT-5.4 Mini leads by +15.7
APEX-Agents · evaluates AI agents on complex, multi-step tasks requiring planning, tool use, and autonomous decision-making in realistic environments.
Claude Haiku 4.5
8.9
GPT-5.4 Mini
24.6
ARC-AGI
GPT-5.4 Mini leads by +16.0
ARC-AGI · the original Abstraction and Reasoning Corpus, testing whether AI can solve novel visual pattern recognition tasks without memorization.
Claude Haiku 4.5
47.7
GPT-5.4 Mini
63.7
ARC-AGI-2
GPT-5.4 Mini leads by +14.9
ARC-AGI-2 · the second iteration of the Abstraction and Reasoning Corpus, testing novel pattern recognition and abstract reasoning without prior training data.
Claude Haiku 4.5
4.0
GPT-5.4 Mini
18.9
Chess Puzzles
GPT-5.4 Mini leads by +16.8
Chess Puzzles · tests strategic and tactical reasoning by having models solve chess puzzle positions, evaluating lookahead and pattern recognition abilities.
Claude Haiku 4.5
3.2
GPT-5.4 Mini
20.0
DeepResearch Bench
Claude Haiku 4.5 leads by +9.2
DeepResearch Bench · evaluates AI on complex multi-step research tasks requiring information gathering, synthesis, and producing comprehensive analyses.
Claude Haiku 4.5
45.5
GPT-5.4 Mini
36.3
Dtbench
GPT-5.4 Mini leads by +10.7
Claude Haiku 4.5
56.0
GPT-5.4 Mini
66.7
FrontierMath-2025-02-28-Private
GPT-5.4 Mini leads by +17.9
FrontierMath (Feb 2025) · original research-level math problems created by mathematicians, testing capabilities at the boundary of current AI mathematical reasoning.
Claude Haiku 4.5
10.4
GPT-5.4 Mini
28.3
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.
Claude Haiku 4.5
2.1
GPT-5.4 Mini
2.1
GPQA diamond
GPT-5.4 Mini leads by +20.9
Graduate-Level Google-Proof QA (Diamond set) · expert-crafted questions in physics, biology, and chemistry that are difficult even for domain PhDs.
Claude Haiku 4.5
61.6
GPT-5.4 Mini
82.5
Lmca
GPT-5.4 Mini leads by +11.6
Claude Haiku 4.5
36.4
GPT-5.4 Mini
48.0
OTIS Mock AIME 2024-2025
GPT-5.4 Mini leads by +22.3
OTIS Mock AIME 2024-2025 · simulated American Invitational Mathematics Examination problems testing advanced problem-solving skills.
Claude Haiku 4.5
66.6
GPT-5.4 Mini
88.9
SimpleQA Verified
GPT-5.4 Mini leads by +16.2
SimpleQA Verified · short factual questions with verified answers, measuring factual accuracy and the tendency to hallucinate or provide incorrect information.
Claude Haiku 4.5
13.2
GPT-5.4 Mini
29.4
WeirdML
GPT-5.4 Mini leads by +14.9
WeirdML · tests models on unusual and adversarial machine learning tasks that require creative problem-solving beyond standard patterns.
Claude Haiku 4.5
45.4
GPT-5.4 Mini
60.3
Full benchmark table
| Benchmark | Claude Haiku 4.5 | GPT-5.4 Mini |
|---|---|---|
APEX-Agents APEX-Agents · evaluates AI agents on complex, multi-step tasks requiring planning, tool use, and autonomous decision-making in realistic environments. | 8.9 | 24.6 |
ARC-AGI ARC-AGI · the original Abstraction and Reasoning Corpus, testing whether AI can solve novel visual pattern recognition tasks without memorization. | 47.7 | 63.7 |
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. | 4.0 | 18.9 |
Chess Puzzles Chess Puzzles · tests strategic and tactical reasoning by having models solve chess puzzle positions, evaluating lookahead and pattern recognition abilities. | 3.2 | 20.0 |
DeepResearch Bench DeepResearch Bench · evaluates AI on complex multi-step research tasks requiring information gathering, synthesis, and producing comprehensive analyses. | 45.5 | 36.3 |
Dtbench | 56.0 | 66.7 |
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. | 10.4 | 28.3 |
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 | 2.1 |
GPQA diamond Graduate-Level Google-Proof QA (Diamond set) · expert-crafted questions in physics, biology, and chemistry that are difficult even for domain PhDs. | 61.6 | 82.5 |
Lmca | 36.4 | 48.0 |
OTIS Mock AIME 2024-2025 OTIS Mock AIME 2024-2025 · simulated American Invitational Mathematics Examination problems testing advanced problem-solving skills. | 66.6 | 88.9 |
SimpleQA Verified SimpleQA Verified · short factual questions with verified answers, measuring factual accuracy and the tendency to hallucinate or provide incorrect information. | 13.2 | 29.4 |
WeirdML WeirdML · tests models on unusual and adversarial machine learning tasks that require creative problem-solving beyond standard patterns. | 45.4 | 60.3 |
Pricing · per 1M tokens · projected $/mo at 10M tokens
| Model | Input | Output | Context | Projected $/mo |
|---|---|---|---|---|
| $1.00 | $5.00 | 200K tokens (~100 books) | $20.00 | |
| $0.75 | $4.50 | 400K tokens (~200 books) | $16.88 |
People also compared