whatmodel.sh / GPT-4.5 Preview
GPT-4.5 Preview
↗ share
+ contribute
GPT-4.5 Preview is a closed-weights language model from OpenAI, released February 2025. Licence: proprietary.
Catalogue identity notes
Comparison model from the Epoch AI internal-evaluation CSVs (Epoch release date 2025-02-27). Dated API snapshot gpt-4.5-preview-2025-02-27. OpenAI deprecations page: gpt-4.5-preview removed 2025-07-14.
$ how it compares
The LLMs ranked either side of GPT-4.5 Preview, out of 828 with a score, then more from OpenAI.
| # | model | score | where | ||
|---|---|---|---|---|---|
| 196 |  Qwen3.5 9B | 63±1 | api $0.11/M +local | ||
| 197 |  Seed-OSS 36B Instruct | 63±1 | local | ||
| 198 | GPT-4.5 Preview | 63±1 | comparison only | ||
| 199 |  Trinity-Large-Thinking | 63±2 | api $0.39/M +local | ||
| 200 |  Magistral Medium 1.2 (2509) | 63±2 | comparison only | ||
| more from OpenAI | |||||
| 1 | GPT-6 Astra | 87±1 | api $20.00/M | ||
| 3 | GPT-6.1 Sol | 85±1 | api $4.00/M | ||
| 7 | GPT-6 Sol | 83±1 | api $4.00/M | ||
$ what it's good at
agent tasks▁▁▂▆█▆▅▄▄▃▂▁ #176 of 340 * · 1.7% of its score · 2 tests
Beats o1 (2024-12-17), o3-mini, DeepSeek R1 0528 here, though they rank higher overall. Trails SmolLM3 3B, Jamba 1.5 Mini, Jamba 1.5 Large here, though they rank lower overall.
| test | result | vs the others in its run | counts |
|---|---|---|---|
| BFCL v3 | 70.3% ✎ | ▁▁▁▁▁▃▅█▅▄▃▁ beats 109/114 | 1.1% |
| Galileo Agent Leaderboard v1 (tool selection quality) | 90.0% ✎ | ▃▃▃▃▆▅▆▅█▂▅▂ beats 22/31 | 0.6% |
coding▁▁▁▃▅▇█▇▅▄▃▁ #177 of 524 · 20% of its score · 3 tests
Beats Gemini 2.5 Pro, Gemini 3.1 Flash-Lite, Qwen3-Coder-Next (80B-A3B) here, though they rank higher overall. Trails Ministral 3 8B Reasoning 2512, Phi-4-reasoning 14B, DeepSeek V3 (Dec '24) here, though they rank lower overall.
| test | result | vs the others in its run | counts |
|---|---|---|---|
| Aider polyglot (225 exercises, two attempts) | 44.9% ✎ | ▂·▂▃▅▂▃█▆▄▂▂ beats 22/49 | 16% |
| WebApp1K-Duo-React pass@1 (onekq-ai) | 66.3% ✎ | ▁··▁·▁▁▁▆█▇█ beats 22/45 | 2.8% |
| HumanEval (simple-evals) lab | 88.6% ✎ | ▆█▆▆▂··▂···▂ beats 6/15 | 1.3% |
chat & writing▁▁▁▃▅▄▆█▇▇▇▃ #75 of 463 · 16% of its score · 3 tests
Beats GPT-5.2, Qwen3.8 27B, LongCat 2.0 (1.6T-A48B) here, though they rank higher overall.
| test | result | vs the others in its run | counts |
|---|---|---|---|
| Arena-Hard v2.0 hard prompts (style control, Gemini-2.5 judge) | 50.0% ✎ | ▂·▃▅▅·▃█▆▂▂▂ beats 14/22 | 8.5% |
| LMArena (text) | 1444.7 ✎ | ▁▁▁▂▃▃▃▆▆▇█▃ beats 307/379 | 5.4% |
| Arena-Hard v2.0 creative writing (GPT-4.1 + Gemini-2.5 judges) | 51.4% ✎ | ▆▂▂▆▂▂▃█▂▃·▂ beats 12/22 | 1.5% |
math▁▂▃▅▇▅▅█▄▁▁▁ #208 of 462 · 18% of its score · 4 tests
Beats MiniMax-M2.1 here, though they rank higher overall. Trails Phi-4-mini-reasoning 3.8B, OpenThinker3 7B, Olmo 3 7B Think here, though they rank lower overall.
| test | result | vs the others in its run | counts |
|---|---|---|---|
| MATH level 5 | 78.6% ✎ | ▂▂▅▆▄█▆▃▅▂▃▄ beats 65/92 | 6.9% |
| OTIS mock AIME 2024–2025 | 37.8% ✎ | ▃▃▄▃▃▃▇▇█▃▂▅ beats 69/193 | 5.5% |
| MATH-500 (simple-evals) lab | 87.1% ✎ | ▃··▅▅▃···▃·█ beats 4/9 | 3.8% |
| MGSM (simple-evals) lab | 86.9% ✎ | ▁····▁▅▁█▁▃▁ beats 2/16 | 1.4% |
vision & documents▁▁▃▃▂▃▅▄▇▇█▃ #76 of 182 · 11% of its score · 3 tests
Beats DeepSeek V4 Flash Vision (Exp), GPT-5, Inkling Small (266B-A12B) here, though they rank higher overall.
| test | result | vs the others in its run | counts |
|---|---|---|---|
| Arena vision (style controlled) | 1225.1 ✎ | ▁▁▂▂▂▂▄▅▅▆██ beats 85/145 | 3.8% |
| CharXiv (reasoning questions) lab | 55.4% ✎ | ▄··▂█··▂···▄ beats 4/9 | 3.6% |
| CharXiv descriptive questions lab | 90.0% ✎ | ▄▄·····▄·▄█▄ beats 6/6 | 3.5% |
other tests8 not in a task · 34% of its score
| test | result | vs the others in its run | counts |
|---|---|---|---|
| GPQA Diamond (simple-evals) lab | 69.5% ✎ | █▃▃█·▃▃▅▃▃▃▅ beats 11/16 | 12% |
| GPQA Diamond | 68.7% ✎ | ▁▁▁▁▂▄▆▄▅▇█▅ beats 103/208 | 10% |
| ARC-AGI-1 public evaluation set (ARC Prize model_baseline, pass@2) | 15.6% ✎ | ▁·▇▃▁·█▇▁·▇▃ beats 6/27 | 4.1% |
| EQ-Bench 3 (emotional intelligence, judge Elo) | 1153.7 ✎ | ▁·▂▁▄▅▅▃█▄▁▁ beats 21/62 | 3.3% |
| MMLU (simple-evals, zero-shot CoT) lab | 90.8% ✎ | ▂▂·█▄▄▂·█▂·▂ beats 13/16 | 3.0% |
| SimpleQA (simple-evals) lab | 62.5% ✎ | ▄·▁▁▁▁··█▁·▃ beats 14/14 | 0.5% |
| DROP F1, 3-shot (simple-evals) lab | 83.4% ✎ | ▂▂▄▆▆█·▂···▄ beats 11/16 | 0.4% |
| Vectara hallucination leaderboard, 2025 dataset (HHEM-2.3, factual consistency rate) | 98.8% ✎ | ▁▁▂▃▄▄█▆▃▂▃▁ beats 134/140 | 0.2% |
$ where to run it
No verified API or download recorded yet · add one
$ how the score was worked out · OpenAI simple-evals README be… 23%, Epoch AI — internal evaluatio… 22%, Aider polyglot benchmark runs 16% and 9 more
Each result implies a score on its own, after adjusting for how hard its test is and who ran it. The score is their weighted average, with a small pull toward the middle for thinly tested models. Each test's weight is its counts column above; here it's added up by who ran the tests. How it works
$ improve this page
Use a short GitHub form or edit the data through a pull request. Every number needs a source; here's which ones we can use.
data updated · sources and licences · json