M
ModelSage
返回
D

DeepSeek V3.1 Terminus (non-reasoning)

DeepSeek · DeepSeek V3.1 Terminus (Non-reasoning)

轻量模型

Input
¥4.0
/M Tokens
Output
¥12.0
/M Tokens
Context
128K
Tokens
Capabilities
对话代码
Best For
简单问答
查看官方定价页面
M
在 ModelSage 接入此模型
Quick Cost Estimate
¥56.0/月
10,000 次调用 · 输入 500 tokens · 输出 300 tokens
自建可行性
vLLM 官方 · 2026-10-14

DeepSeek V3.1 Terminus (non-reasoning) 私有化部署，NVFP4 量化下需显存约 403GB， 至少 3 张 H200，聚合吞吐约 14,033 tok/s。

NVFP4403GB
3×H200 / 2×MI325X
FP8805GB
6×H200 / 3×MI355X
vLLM 官方验证过的硬件
H200MI300XMI325XMI355X

同样的量走 API：日均 2 亿 token 约 ¥36,000/月。自建是否更便宜取决于卡价与实际利用率。

按实际卡价测算回本周期

显存需求与硬件兼容性取自 vLLM 官方 recipes；卡数按「装得下」估算， 未计入张量并行度、互联拓扑与跨节点开销，实际只多不少。吞吐为公式估算值，非实测。

Leaderboard Ranks
🏅综合#252💻代码#148
Price Comparison同价位的其他厂商，点进去看详情
DeepSeek V3.1 Terminus (non-reasoning)CURRENT
¥4.0 / ¥12.0
INPUT
OUTPUT
ERNIE 4.5 300B A47B
¥4.0 / ¥16.0
INPUT
OUTPUT
Kimi K2.5
¥4.0 / ¥21.0
INPUT
OUTPUT
Kimi K2 Thinking
¥4.0 / ¥16.0
INPUT
OUTPUT
Kimi K2.5 (non-reasoning)
¥4.0 / ¥21.0
INPUT
OUTPUT
DeepSeek Models
DeepSeek V4.1 Flash (max)
¥2.0 / ¥8.0
DeepSeek V4 Pro 0813 (max)
¥9.0 / ¥27.0
DeepSeek V4 Flash Vision (max)
¥2.9 / ¥8.8
DeepSeek V4 Flash 0731 (max)
¥2.0 / ¥8.0
DeepSeek V4 Pro (max)
¥9.0 / ¥27.0
DeepSeek V4 Pro (high)
¥9.0 / ¥27.0
DeepSeek V4.1 Flash (non-reasoning)
¥2.0 / ¥8.0
DeepSeek V4 Flash (high)
¥2.0 / ¥8.0
DeepSeek V4 Flash (max)
¥2.0 / ¥8.0
DeepSeek V3.2
¥2.0 / ¥3.0
DeepSeek V4 Pro (non-reasoning)
¥9.0 / ¥27.0
DeepSeek V4 Pro 0813 (non-reasoning)
¥9.0 / ¥27.0
DeepSeek V4 Flash (non-reasoning)
¥2.0 / ¥8.0
DeepSeek V3.2 Exp
¥2.0 / ¥3.0
DeepSeek V3.2 (non-reasoning)
¥2.0 / ¥3.0
DeepSeek V3.1 Terminus
¥4.0 / ¥12.0
DeepSeek V3.2 Exp (non-reasoning)
¥2.0 / ¥3.0
DeepSeek V3.1 (non-reasoning)
¥4.0 / ¥12.0
DeepSeek V3.1
¥4.0 / ¥12.0
DeepSeek R1 0528
¥4.0 / ¥16.0
DeepSeek R1 (Jan)
¥4.0 / ¥16.0
DeepSeek V3 0324
¥2.0 / ¥8.0
DeepSeek V3 (Dec)
¥2.0 / ¥8.0
DeepSeek R1 Distill Llama 70B
¥4.0 / ¥16.0
LAST UPDATED: 2026-10-14