DeepSeek releases V3.1-Terminus open-weight model
What happened
On October 3, DeepSeek released the open-weight model DeepSeek-V3.1-Terminus. It builds on V3.1 with improvements based on user feedback. The report also discloses three evaluation results: 68.2 on LMCA, 99% on DTBench and 2951 on ALE-Bench. Pricing is $0.27 per million input tokens and $1 per million output tokens. The disclosed development covers the release, evaluation scores and pricing; the report does not describe the specific improvements.
Written by AI from the coverage · updated 2 hours ago
Coverage
Follow the reports to see the story from different sides.
- AI HOT · ModelsDeepSeek 发布开源权重模型 DeepSeek-V3.1-Terminus
DeepSeek 发布开源权重模型 DeepSeek-V3.1-Terminus，基于 V3.1 并针对用户反馈做了改进。该模型在 LMCA 上得分 68.2、DTBench 达 99%、ALE-Bench 达 2951，定价为输入 $0.27 / 1M tokens、输出 $1 / 1M tokens。
Heat over time
Not enough continuous observations to draw a trend yet.