Estimate from the version's published specs, benchmarks and reviews, 0-5; scores are only comparable within one criteria year. Scored as launched (gpt-3.5-turbo-0125, 25 Jan 2024): 16,385-token context (OpenAI model page, SiliconANGLE '16k context'); $0.50 in / $1.50 out per 1M tokens = $0.75 blended (The Decoder). Text in, text out. No AA Intelligence Index, Terminal-Bench or maker/AA speed figure at launch, so reasoning, coding, speed null (a forum user measured 82 tokens/s on short outputs; not scored, not a provider or AA figure). Reviews: two news write-ups plus two OpenAI Developer Community threads of developers' own tests (mixed: faster, but one reports regressions vs 0613); no independent tech-press hands-on review of this minor update found. openai.com post blocked (403); sourceUrl is OpenAI's model page.[3]
Input prices have been reduced by 50 percent to $0.0005 per 1K token and output prices have been reduced by 25 percent to $0.0015 per 1K token.
improvements in instruction following, more reproducible outputs and support parallel function calling
It also appears fast compared to 1106.
I clearly notice that gpt-3.5-turbo-0125 is worst than gpt-3.5-turbo-0613 (and previous ones), I do notice that in many situations, it seems to output the less possible.
| OpenAI ships less lazy GPT-4 model and lowers prices for GPT-3.5 · 2024年1月25日 | 0.30 |
|---|---|
| OpenAI offers new embedding models, updates GPT-4 and GPT-3.5, and lowers API access cost · 2024年1月25日 | 0.40 |
| Gpt-3.5-turbo-0125 is available via API now. Share your first impressions · 2024年2月2日 | 0.00 |
| Gpt-3.5-turbo-0125 worst than gpt-3.5-turbo-0613 · 2024年2月25日 | -0.40 |
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[2][3][4] 于 2024 年 1 月 25 日宣布 gpt-3.5-turbo-0125 将在“next week”到来;OpenAI API 更新日志中 “Feb 1” 的条目写道 “Released gpt-3.5-turbo-0125, an updated GPT-3.5 Turbo model”。
OpenAI's[1][3][4] API changelog: 6 Nov 2023 updated GPT-3.5 Turbo, 25 Jan 2024 embedding V3 models and updated GPT-4 Turbo preview, and 1 Feb 2024 'Released gpt-3.5-turbo-0125'.
Current model page: default snapshot gpt-3.5-turbo-0125, 16,385 context window, 4,096 max output tokens, September 2021 knowledge cutoff, $0.5 input and $1.5 output per 1 million tokens, and the advice to use gpt-4o-mini instead since July 2024.[1]
Lists gpt-3.5-turbo-0125 and the gpt-3.5-turbo alias for shutdown on 23 October 2026, with gpt-5.6-terra as the substitute.[1]
Snapshot of OpenAI's[2][3][4] 25 January 2024 post, which openai.com[1] blocks to scripts: the new gpt-3.5-turbo-0125 'next week', input down 50% to $0.0005 and output down 25% to $0.0015 per 1K tokens, the format-following and non-English function-call fixes, and the alias upgrade two weeks after launch.
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-3.5 Turbo 0125 Released