Resources

View closed (9)

Distill the model into 4 variants: 600M, 1.5B, 4B and 7B

#48 opened 17 days ago by

ShreyashSarkarDev

模型在复杂工具调用出现输出乱码的

#47 opened about 1 month ago by

lichenxi2222

Quantum Ai

#45 opened about 1 month ago by

Xavernox

调用不存在的 tool

#43 opened about 2 months ago by

Jingxin

中文 markdown 表格容易出错

#42 opened about 2 months ago by

wunengbiao

OUTPUT_PARSING_FAILURE

#41 opened about 2 months ago by

jinysun

RuntimeError: No DeepseekV2MoE layer found in model.layers.

#40 opened about 2 months ago by

ajtakto

Script to download deepseek tensor files

#39 opened about 2 months ago by

gariandamo

Criticism, please read

👍 ❤️ 9

#38 opened about 2 months ago by

ChuckMcSneed

Solve this Isuue Please

#37 opened about 2 months ago by

deep7654

多轮tool call能力不稳定

#35 opened 2 months ago by

xielegend

Have you guys noticed this "极""極""extreme"problem?

➕ 6

#34 opened 2 months ago by

DingzhenPearl

为什么模型卡片里的weight_scale_inv即使用fp32存储，也还是存在尾数？

#33 opened 2 months ago by

Ali1234588

Feat: Add toolcall description

#32 opened 2 months ago by

eric8810

Update README.md

#29 opened 2 months ago by

AmeelF

[W4A8 FP8 Quantization] Release of DeepSeek-V3.1 with SGLang Support – Near-Lossless & 1.56x Speed Boost!

🔥 2

#28 opened 2 months ago by

Carson

tool call for reasoning mode

➕ 5

#27 opened 2 months ago by

shing3232

V3.1 seems to be pretty bad at everything except coding and mathematics. V3.1 看起来除了编程和数学之外，其他方面都很差。

👀 ➕ 9

#26 opened 2 months ago by

qazqazqazqaz46

DeepSeek-V3.1全方位最新实测出炉（300+维度），欢迎进群交流讨论~

#25 opened 2 months ago by

JEIN

有人尝试过本地通过vllm部署并连接到claude code吗？

#24 opened 2 months ago by

Yuxin362

why all the scale values of attn out_proj have the none-zero mantissa 0b111010101010010101001111111101 ?

#22 opened 2 months ago by

abcstar

Update system prompt to include tools

#21 opened 2 months ago by

bchenfireworks

recommended temp?

#19 opened 2 months ago by

createthis

Tool calling usage examples

#18 opened 2 months ago by

1000Xia

Context length: is it 128K (as mentioned in the model card) or 160K (as specified in config.json)?

#17 opened 2 months ago by

Lissanro

不知道还有没有蒸馏模型

#13 opened 2 months ago by

BlackLeee

no score on swe under thinking mode

➕ 👀 3

#11 opened 2 months ago by

vitvamer

This model’s censorship is insane

👍 🧠 6

#10 opened 2 months ago by

smile1030

请问simpleQA结果是否为笔误？

#9 opened 2 months ago by

rnc000

梁文疯垃圾模型

😎 🧠 3

#7 opened 2 months ago by

eiskalt

search是用think模式吗

#6 opened 2 months ago by

awdrgyjilplij

Congratulations to DeepSeek, this version seems powerful for coding and agent development

🔥 ❤️ 1

#5 opened 2 months ago by

Robin-Han

火速下载

#4 opened 2 months ago by

HowardChenRV

Any plan to release the post-training recipe?

#3 opened 2 months ago by

Yi30

Come on, third party bros! Deploy it!

👍 🔥 6

#2 opened 2 months ago by

DingzhenPearl

火前留名

#1 opened 2 months ago by

loveaim