DeepSeek-V4-Pro-0813 · Terminal-Bench(2.1; metric not printed in V4-0813/0731 cards) 87.9 · 主文证据(ev-v4-pro-0813-001)

DeepSeek-V4-Pro-0813 的 Terminal-Bench · 2.1; metric not printed in V4-0813/0731 cards 官方出处。DeepSeek-V4-Pro-0813, max reasoning effort · 87.9。见《DeepSeek-V4-Pro-0813 official Hugging Face model card》Official model card / table headed “| Benchmark | DeepSeek-V4-Pro-0813 | DeepSeek-V4-Flash-0731 | DeepSeek-V4-Pro (Preview) | DeepSeek-V4-Flash (Preview) | GLM-5.2 | Kimi K3 | Opus-4.8 | Fable-5 (w/ fallback) |” / complete row containing “| Terminal Bench 2.1 | 87.9 | 82.7 | 72.1 | 61.8 | 81.0 | 88.3 | 85.0 | 88.0 |”。证据 ev-v4-pro-0813-001。

原文与定位

DeepSeek-V4-Pro-0813 · Terminal-Bench(2.1; metric not printed in V4-0813/0731 cards),DeepSeek-V4-Pro-0813, max reasoning effort配置,Reported score 87.9(具体成绩公布时间未确认)。见《DeepSeek-V4-Pro-0813 official Hugging Face model card》Official model card / table headed “| Benchmark | DeepSeek-V4-Pro-0813 | DeepSeek-V4-Flash-0731 | DeepSeek-V4-Pro (Preview) | DeepSeek-V4-Flash (Preview) | GLM-5.2 | Kimi K3 | Opus-4.8 | Fable-5 (w/ fallback) |” / complete row containing “| Terminal Bench 2.1 | 87.9 | 82.7 | 72.1 | 61.8 | 81.0 | 88.3 | 85.0 | 88.0 |”。 查看成绩条件

| Benchmark | DeepSeek-V4-Pro-0813 | DeepSeek-V4-Flash-0731 | DeepSeek-V4-Pro (Preview) | DeepSeek-V4-Flash (Preview) | GLM-5.2 | Kimi K3 | Opus-4.8 | Fable-5 (w/ fallback) | | Terminal Bench 2.1 | 87.9 | 82.7 | 72.1 | 61.8 | 81.0 | 88.3 | 85.0 | 88.0 |
官方材料
DeepSeek-V4-Pro-0813 official Hugging Face model card
定位
主文 · Official model card / table headed “| Benchmark | DeepSeek-V4-Pro-0813 | DeepSeek-V4-Flash-0731 | DeepSeek-V4-Pro (Preview) | DeepSeek-V4-Flash (Preview) | GLM-5.2 | Kimi K3 | Opus-4.8 | Fable-5 (w/ fallback) |” / complete row containing “| Terminal Bench 2.1 | 87.9 | 82.7 | 72.1 | 61.8 | 81.0 | 88.3 | 85.0 | 88.0 |”
材料版本
HF revision 72e1d3230f6c080a530b0a1d46f8eb4602340597; current card snapshot
版本公布
未确认
披露时间
未确认 · 首发/后补未确认
采集时间
历史依据
暂无可靠历史版本依据
引用文本

DeepSeek-V4-Pro-0813 · Terminal-Bench(2.1; metric not printed in V4-0813/0731 cards),DeepSeek-V4-Pro-0813, max reasoning effort配置,Reported score 87.9(具体成绩公布时间未确认)。官方出处 DeepSeek-V4-Pro-0813 official Hugging Face model card(主文,HF revision 72e1d3230f6c080a530b0a1d46f8eb4602340597; current card snapshot)Official model card / table headed “| Benchmark | DeepSeek-V4-Pro-0813 | DeepSeek-V4-Flash-0731 | DeepSeek-V4-Pro (Preview) | DeepSeek-V4-Flash (Preview) | GLM-5.2 | Kimi K3 | Opus-4.8 | Fable-5 (w/ fallback) |” / complete row containing “| Terminal Bench 2.1 | 87.9 | 82.7 | 72.1 | 61.8 | 81.0 | 88.3 | 85.0 | 88.0 |”,https://huggingface.co/deepseek-ai/DeepSeek-V4-Pro-0813。Benchlog https://benchlog.lekshome.cc/evidence/deepseek-v4-pro-0813/ev-v4-pro-0813-001/

关联成绩与条件

DeepSeek-V4-Pro-0813 · Terminal-Bench 成绩与配置
指标与成绩配置与评测条件证据关系
Reported score87.9原文 87.9 · 单位 未标注
公布 未确认
DeepSeek-V4-Pro-0813, max reasoning effort
设置
The card identifies Max reasoning effort and DeepSeek Harness Minimal mode for code-agent evaluations; temperature=1.0 and top_p=0.95.
未知项
Effort and harness applicability is only asserted for the card-defined code-agent/public rows.
工具
工具条件未记录
推理预算
max
框架
DeepSeek Harness Minimal
协议
For code-agent public benchmark rows, card note: DeepSeek Harness Minimal, max reasoning effort, temperature=1.0, top_p=0.95.
条件
The card does not print a metric for this row. It is retained as a reported score without merging with a Pass@1 or other metric.

直接支持该成绩

在排名中查看此成绩