| 输入 | 文本、图像 |
|---|
| API 价格(每 100 万 tokens,输入 / 缓存 / 输出) | $75 / $37.5 / $150 |
|---|
来自 OpenAI API 模型页面(该模型已不再是当前版本);页面未公布基准测试数据,因此未保存图表。 · 来源
Estimate from the version's published specs, benchmarks and reviews, 0-5; scores are only comparable within one criteria year. Scored as launched (research preview, 27 Feb 2025): 128,000-token context and text+image input (Willison); $75 in / $150 out per 1M tokens = $93.75 blended, over the $40 anchor -> 0 (TechCrunch, Willison). No AA Intelligence Index entry or Terminal-Bench run opened; TechCrunch notes it trails reasoning models on AIME/GPQA and roughly matches GPT-4o/o3-mini on SWE-bench Verified (OpenAI-published, no figure scored), so reasoning and coding null. Speed null: Willison saw a 112 s response for one SVG task, anecdotal only. openai.com post blocked (403); sourceUrl is OpenAI's model page for gpt-4.5-preview. Fetched via developers.openai.com. Reviews are news plus commentary; none is a controlled test. developers.openai.com
Accessed via the API the model feels weirdly slow
On several AI benchmarks, GPT-4.5 comes under newer AI "reasoning" models from Chinese AI startup DeepSeek, Anthropic, and OpenAI itself.
GPT-4.5 is a "giant, expensive model," Altman wrote in Thursday's X post.
GPT-4.5 doesn't bring any groundbreaking advancements in reasoning, problem-solving, or handling complex tasks.
The incremental improvement does not justify the cost.
| Initial impressions of GPT-4.5 · 2025年2月27日 | -0.30 |
|---|---|
| OpenAI unveils GPT-4.5 'Orion,' its largest AI model yet · 2025年2月27日 | -0.30 |
| OpenAI launching GPT-4.5, its next general-purpose large language model · 2025年2月27日 | -0.20 |
| GPT4.5: A Complete Review and How It Compares To Others · 2025年2月28日 | -0.50 |
| Is GPT-4.5 Worth the Hype? · 2025年3月1日 | -0.50 |
读者上传的照片,与图钉带来源的媒体分开显示。你可以为吃到或看到的内容评分,在餐厅图钉上还能直接在照片上标记菜品。
第一项始终是图钉的来源。总体可信度是各份资料对上方所用开始和结束时间支持程度的加权平均;资料每比最新的一份旧 180 天,权重减半。
OpenAI 官方的模型发布说明将该条目日期标为2025年2月27日
States GPT-4.5 (Orion) was officially released February 27, 2025 to Plus and Pro.[1]
CNBC report published on launch day about GPT-4.5 rollout.
Archived copy of OpenAI's[1] GPT-4.5 announcement, which says it "was trained on Microsoft Azure AI supercomputers", gives SimpleQA accuracy of 62.5% and hallucination rate of 37.1%, and lists search, file and image uploads and canvas but not Voice Mode.
有遗漏或错误吗?用你自己的话说明:能佐证此图钉的链接、不同的开始或结束日期及理由,或缺失、有误的信息。AI 会对照此图钉的来源进行核实,搜索更好的来源,并添加任何支持你说法的页面。图钉自身的来源仍然最重要。AI 也会查看图片:显示的是别的东西或显示效果差的图片会被移到后面或替换。
原标题: GPT-4.5 Released as a Research Preview