Perplexity 发布 Decider v1.3 与 Decider Flash v1.3 决策模型
Perplexity 更新了自己的决策模型线,Flash 版从 60 分涨到 64 分,4B 小模型却进了前十。两款都开源且输入价格低到 $0.01-0.02/百万 token,做任务路由或工具调用选型的人可以直接对比成本榜。
Perplexity 更新了 v1.2 决策模型,发布 Decider v1.3 和 Decider Flash v1.3,在 Artificial Analysis Decision Index v2.1 上分别得到 70 分和 64 分。Decider v1.3 保持第 6 名,Decider Flash v1.3 升至第 10 名。
具体来看,Decider v1.3(27B)在 AA-Paperclip 上以 69 分排第 4,比 v1.2 提升 2 分;在 EvalSafe-O*NET-AA 上以 89 分排第 3,比 v1.2 提升 1 分。Decider Flash v1.3(4B)比 v1.2 提升 4 分,主要来自 AA-Symmetry 和 When2Call 各 12 分的提升。
两款模型都位于能力-成本 Pareto 前沿,每百万任务成本分别为 $68 和 $34,输入价格分别为 $0.02 和 $0.01 每百万 token。支持 256K 上下文、文本和图像输入,采用 Apache License 2.0 开源,通过 Perplexity 一方 API 提供。
Decider v1.3 and Decider Flash v1.3 from Perplexity score 70 and 64, respectively, on the Artificial Analysis Decision Index v2.1
@perplexity_ai has released Decider v1.3 and Decider Flash v1.3, updating their v1.2 flagship decision models. Compared to v1.2, Decider v1.3 maintains its #6 position while Decider Flash v1.3 moves up into the #10 position.
Key results:
➤ Decider v1.3 has strong decision-making and extraction capabilities. Decider v1.3 is the #4 model on AA-Paperclip with a score of 69. It improves 2 points over Decider v1.2. It is also the #3 model on EvalSafe-O*NET-AA with a score of 89, a 1-point improvement over Decider v1.2
➤ Decider Flash v1.3 is more capable than its v1.2 predecessor. Decider Flash v1.3 achieves a score of 64 on the Artificial Analysis Decision Index v2.1, a 4-point improvement over v1.2 driven by a 12-point improvement on both AA-Symmetry and When2Call
➤ Decider v1.3 and Decider Flash v1.3 sit on the Capability vs. Cost Pareto frontier. With a Cost per 1M Tasks of $68 for Decider v1.3 and $34 for Decider Flash v1.3, both models occupy positions on the Capability vs. Cost Pareto frontier
Additional model details:
➤ Context window: 256K tokens
➤ Pricing: $0.02 and $0.01 per 1M input tokens for Decider v1.3 and Decider Flash v1.3, respectively
➤ Multimodality: Text and image input
➤ Size: 27B and 4B for Decider v1.3 and Decider Flash v1.3, respectively
➤ License: Apache License 2.0
➤ Providers: Available through Perplexity’s first-party API
本条情报归属 AITOP「模型」核心栏目,重点聚焦北美前沿实验室与硅谷科技巨头。在跨区域产业观察坐标中,该动态直接关联「AI 开发者工具链、Coding Agent 与工程化实践」「中美大模型技术代差与商业化对照」研究主线,反映出当前技术演进与商业落地的最新风向。涉及核心实体与技术议题包括:Perplexity、Decider、开源模型、决策模型,读者可通过下方标签进入对应专题页查阅历史脉络与关联对比。
信源核验说明:本文原始线索采编自「Artificial Analysis」(@ArtificialAnlys),经 AITOP 多维评分管线结构化提炼。