GPT-4o: benchmark scores, price and ranking

GPT-4o from OpenAI ranks #458 of 743 LLMs on the LLMs Tiger aggregate benchmark score (47.2), based on 6 benchmark results captured up to 2026-10-13.

Organisation: OpenAI
Type: Frontier model
Open weights: No
Released: 2024-05-13 (source: curated)
Context window: 128,000 tokens
Cheapest API price: $2.50 in / $10.00 out per million tokens (via azure, observed 2026-10-13T20:47Z)
Aggregate rank: #458 of 743 (score 47.2)

Relative to other models, GPT-4o is strongest on Aider Refactoring (#5 of 12) and weakest on ARC-AGI-2 (#81 of 87).

Across the 5 benchmarks where at least ten models are scored, GPT-4o is in the top 10% on 0 and in the top half on 3.

Its cheapest listed input price, $2.50 per million tokens, is lower than that of 15% of the 427 models with a listed price.

GPT-4o benchmark results

GPT-4o benchmark scores, ranked against all models measured on each benchmark
Benchmark	Score	Rank	Source	Captured
Aider Refactoring	62.9%	#5 of 12	Aider Aider Refactoring leaderboard	2026-10-13
SWE-bench Test	14.6%	#5 of 11	SWE-bench Test leaderboard	2026-10-13
SWE-bench Lite	24.7%	#10 of 20	SWE-bench Lite leaderboard	2026-10-13
SWE-bench Multimodal	30.4%	#3 of 5	SWE-bench Multimodal leaderboard	2026-10-13
SWE-bench Verified	26.2%	#62 of 73	SWE-bench Verified leaderboard	2026-10-13
ARC-AGI-2	0.0%	#81 of 87	ARC Prize official leaderboard	2026-10-13

About GPT-4o

GPT-4o is a frontier model from OpenAI. It was released on 2024-05-13, according to the developer's announcement. It accepts up to 128,000 tokens of input at once. The cheapest API listing we track charges $2.50 per million input tokens and $10.00 per million output tokens, through azure. We hold 6 benchmark results for it across 3 areas, and none is in the top tenth of its board.

Where GPT-4o stands

On our aggregate score, GPT-4o is #458 of 743 ranked models, in the lower half. Its score is 47.2; the scale is centred on 50, so it sits 2.8 points below the typical model. That beats 38% of the ranked field. Among the 88 ranked OpenAI models it is #59. The next one up is gpt-5.6-luna-xhigh (codex-harness) at 47.2. The next one down is GPT-4o mini at 47.1.

GPT-4o, benchmark by benchmark

Aider Refactoring puts GPT-4o at 62.9%, #5 of 12. That beats 58% of the models tested, but 29.2 points short of Claude 3.5-Sonnet-20241022 (92.1%). The results either side are claude-3.5-sonnet-20240620 (64.0%) and gpt-4-1106-preview (50.6%). It covers Aider official benchmark. The board median is 50.0%, so GPT-4o is 12.9 points above it. Source: Aider Aider Refactoring leaderboard, captured 2026-10-13.

On SWE-bench Test, GPT-4o reaches 14.6%. It ranks #5 of 11, ahead of 55% of the field, and 38.0 points behind Claude 4.5 Opus at 52.6%. The results either side are Claude 3.5 Sonnet (18.1%) and GPT-4 (1106) (12.5%). It covers real GitHub issues resolved. The board median is 12.5%, so GPT-4o is 2.1 points above it. Source: SWE-bench Test leaderboard, captured 2026-10-13.

GPT-4o manages 24.7% on SWE-bench Lite, good for #10 of 20. Claude 4 Sonnet leads at 56.7%, 32.0 points further up, and 50% of models sit below. The results either side are CodeAct v1.8 (26.7%) and Qwen2.5 (7B + 72B) (24.7%). It covers real GitHub issues resolved. The median score on this board is also 24.7%. Source: SWE-bench Lite leaderboard, captured 2026-10-13.

SWE-bench Verified is a weak spot. GPT-4o scores 26.2%, #62 of 73, behind 84% of the models measured. The leader, Claude Opus 4.7 (max), reaches 83.5%. It covers real GitHub issues resolved. The board median is 56.4%, so GPT-4o is 30.2 points below it. Source: SWE-bench Verified leaderboard, captured 2026-10-13.

ARC-AGI-2 is a weak spot. GPT-4o scores 0.0%, #81 of 87, behind 92% of the models measured. The leader, GPT-6 Astra (max), reaches 95.0%. It covers abstract reasoning, ARC Prize public eval set. The board median is 22.8%, so GPT-4o is 22.8 points below it. Source: ARC Prize official leaderboard, captured 2026-10-13.

The full table above also lists the other 1 result, including boards with fewer than ten models, where a rank says little.

Strengths and weaknesses

Software engineering: uneven, ahead of 33% of the field on average across 2 benchmarks. Its best board here is SWE-bench Lite (#10 of 20) and its weakest is SWE-bench Verified (#62 of 73). General knowledge and chat: solid, ahead of 56% of the field on average across 2 benchmarks. Its best board here is Aider Refactoring (#5 of 12) and its weakest is SWE-bench Test (#5 of 11).

Price and value

The median input price across the 427 priced models is $0.500 per million tokens, so GPT-4o costs more than the typical listing. Models that score within 1.5 points or better and cost less: Qwen2.5-Coder-7B-Instruct ($0.010, 49.5), DeepSeek V4.1 Flash (high) ($0.010, 48.9), Qwen2.5-Coder-3B-Instruct ($0.010, 46.5). Similar scores cost more on Grok 4 (Fast Reasoning) ($3.00) and grok-3 ($3.00).

Release history

The OpenAI model released before it was gpt-4-turbo-2024-04-09 on 2024-04-09, which scores 42.6. The next one was gpt-4o-2024-05-13 on 2024-05-13, at 51.9.

Reading these numbers

The 6 scores on this page come from 6 sources: ARC Prize official leaderboard (1), Aider Aider Refactoring leaderboard (1), SWE-bench Lite leaderboard (1), SWE-bench Multimodal leaderboard (1). 6 of them come from an official leaderboard and 0 from datasets or third-party runs. With 6 results the aggregate rank can move a lot when one new score arrives.