GPT-4.5 Preview (Feb 2025): benchmark scores, price and ranking
GPT-4.5 Preview (Feb 2025) from OpenAI ranks #501 of 743 LLMs on the LLMs Tiger aggregate benchmark score (46.6), based on 8 benchmark results captured up to 2026-10-13.
- Organisation
- OpenAI
- Type
- Preview release
- Open weights
- No
- Released
- 2025-02-27 (source: epoch)
- Cheapest API price
- $75.00 in / $150.00 out per million tokens (via azure, observed 2026-10-13T19:47Z)
- Aggregate rank
- #501 of 743 (score 46.6)
Relative to other models, GPT-4.5 Preview (Feb 2025) is strongest on Fiction.liveBench (#21 of 47) and weakest on Humanity's Last Exam (#20 of 27).
Across the 8 benchmarks where at least ten models are scored, GPT-4.5 Preview (Feb 2025) is in the top 10% on 0 and in the top half on 2.
Its cheapest listed input price, $75.00 per million tokens, is lower than that of 0% of the 427 models with a listed price.
GPT-4.5 Preview (Feb 2025) benchmark results
| Benchmark | Score | Rank | Source | Captured |
|---|---|---|---|---|
| Fiction.liveBench | 63.9% | #21 of 47 | Epoch AI Benchmarking Hub | 2026-10-13 |
| EnigmaEval | 3.2% | #11 of 24 | Epoch AI Benchmarking Hub | 2026-10-13 |
| GPQA Diamond | 68.7% | #146 of 275 | Epoch AI Benchmarking Hub | 2026-10-13 |
| Cybench | 17.5% | #9 of 16 | Epoch AI Benchmarking Hub | 2026-10-13 |
| WeirdML | 39.4% | #63 of 109 | Epoch AI Benchmarking Hub | 2026-10-13 |
| OTIS Mock AIME | 37.8% | #116 of 189 | Epoch AI Benchmarking Hub | 2026-10-13 |
| SimpleBench | 34.5% | #39 of 61 | Epoch AI Benchmarking Hub | 2026-10-13 |
| Humanity's Last Exam | 5.4% | #20 of 27 | Epoch AI Benchmarking Hub | 2026-10-13 |
About GPT-4.5 Preview (Feb 2025)
GPT-4.5 Preview (Feb 2025) is a preview release from OpenAI. It was released on 2025-02-27, according to Epoch AI. The cheapest API listing we track charges $75.00 per million input tokens and $150.00 per million output tokens, through azure. We hold 8 benchmark results for it across 2 areas, and none is in the top tenth of its board.
Where GPT-4.5 Preview (Feb 2025) stands
On our aggregate score, GPT-4.5 Preview (Feb 2025) is #501 of 743 ranked models, in the lower half. Its score is 46.6; the scale is centred on 50, so it sits 3.4 points below the typical model. That beats 33% of the ranked field. Among the 88 ranked OpenAI models it is #68. The next one up is o4-mini (high) at 46.8. The next one down is GPT-5.4 mini (xhigh) at 46.3.
GPT-4.5 Preview (Feb 2025), benchmark by benchmark
Fiction.liveBench puts GPT-4.5 Preview (Feb 2025) at 63.9%, #21 of 47. That beats 55% of the models tested, but 30.5 points short of grok-4-0709 (94.4%). The results either side are GPT-4.1 (63.9%) and qwen3-14b (62.5%). It covers long-context story comprehension (16k). The board median is 61.1%, so GPT-4.5 Preview (Feb 2025) is 2.8 points above it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.
On EnigmaEval, GPT-4.5 Preview (Feb 2025) reaches 3.2%. It ranks #11 of 24, ahead of 54% of the field, and 15.1 points behind Gemini 3 Pro Preview at 18.2%. The results either side are Kimi K2.5 (3.4%) and Claude Sonnet 4 (unknown thinking) (3.1%). It covers multi-step puzzle hunts. The board median is 2.9%, so GPT-4.5 Preview (Feb 2025) is 0.3 points above it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.
GPT-4.5 Preview (Feb 2025) lands below the middle on GPQA Diamond: 68.7%, #146 of 275, with 53% of the field scoring higher. GPT-6 Astra (max) leads at 95.8%; the closest result above is GPT-5 nano (high) at 69.4%. It covers expert-level science questions. The board median is 71.7%, so GPT-4.5 Preview (Feb 2025) is 3.0 points below it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.
GPT-4.5 Preview (Feb 2025) lands below the middle on Cybench: 17.5%, #9 of 16, with 50% of the field scoring higher. grok-4-0709 leads at 43.0%; the closest result above is Claude 3.5 Sonnet (Jun 2024) at 17.5%. It covers cybersecurity capture-the-flag tasks. The median score on this board is also 17.5%. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.
GPT-4.5 Preview (Feb 2025) lands below the middle on WeirdML: 39.4%, #63 of 109, with 57% of the field scoring higher. GPT-6 Astra (max) leads at 93.3%; the closest result above is Qwen3.5 27B at 39.5%. It covers unusual machine-learning tasks. The board median is 41.2%, so GPT-4.5 Preview (Feb 2025) is 1.8 points below it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.
GPT-4.5 Preview (Feb 2025) lands below the middle on OTIS Mock AIME: 37.8%, #116 of 189, with 61% of the field scoring higher. Claude Fable 5.1 (max) leads at 100.0%; the closest result above is DeepSeek-V3 (Mar 2025) at 37.8%. It covers competition maths (AIME 2024-25 style). The board median is 57.8%, so GPT-4.5 Preview (Feb 2025) is 20.0 points below it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.
GPT-4.5 Preview (Feb 2025) lands below the middle on SimpleBench: 34.5%, #39 of 61, with 62% of the field scoring higher. Claude Fable 5 (max) leads at 81.9%; the closest result above is Qwen 3.6 Flash (2026-04-16) at 35.2%. It covers trick questions that test common sense. The board median is 41.4%, so GPT-4.5 Preview (Feb 2025) is 6.9 points below it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.
GPT-4.5 Preview (Feb 2025) lands below the middle on Humanity's Last Exam: 5.4%, #20 of 27, with 70% of the field scoring higher. Muse Spark leads at 40.6%; the closest result above is Llama 4 Maverick at 5.7%. It covers expert-level questions across domains. The board median is 10.7%, so GPT-4.5 Preview (Feb 2025) is 5.3 points below it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.
Strengths and weaknesses
Reasoning, maths and science: uneven, ahead of 40% of the field on average across 5 benchmarks. Its best board here is EnigmaEval (#11 of 24) and its weakest is Humanity's Last Exam (#20 of 27). General knowledge and chat: uneven, ahead of 47% of the field on average across 3 benchmarks. Its best board here is Fiction.liveBench (#21 of 47) and its weakest is WeirdML (#63 of 109).
Price and value
The median input price across the 427 priced models is $0.500 per million tokens, so GPT-4.5 Preview (Feb 2025) 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).
Release history
The OpenAI model released before it was GPT-4.5 on 2025-02-27, which scores 47.5. The next one was chatgpt-4o-latest (2025-03-29) on 2025-03-29, at 50.2.
Reading these numbers
The 8 scores on this page come from 1 source: Epoch AI Benchmarking Hub (8). No official leaderboard supplied any of them, so treat them as third-party measurements.
Models ranked nearby
- #498 gemini-2.0-flash-lite (46.6)
- #499 Mellum2-12B-A2.5B-Thinking-SFT (46.6)
- #500 EXAONE-3.5-7.8B-Instruct (46.6)
- #502 mimo-v2-pro (46.6)
- #503 Yi-1.5-9B-Chat (46.6)
- #504 Gemini Robotics-ER 1.6 Preview (46.6)
More from OpenAI
- GPT-6 Astra (max) (#2, 64.0)
- GPT-6.1 Sol (max) (#5, 60.6)
- GPT-5.6 Sol (max) (#6, 60.1)
- GPT-6 Sol (max) (#7, 60.0)
- GPT-5.5 (xhigh) (#11, 58.0)
- GPT-5.6 Terra (max) (#12, 57.8)
Compare it on the live leaderboard → · How the aggregate score works