跳轉到主要內容

DeepSeek R1 Zero

DeepSeekDeepSeek開源權重MIT · 商用許可

描述

DeepSeek-R1-Zero, a model trained via large-scale reinforcement learning (RL) without supervised fine-tuning (SFT) as a preliminary step, demonstrated remarkable performance on reasoning. With RL, DeepSeek-R1-Zero naturally emerged with numerous powerful and interesting reasoning behaviors. However, DeepSeek-R1-Zero encounters challenges such as endless repetition, poor readability, and language mixing. To address these issues and further enhance reasoning performance, we introduce DeepSeek-R1, which incorporates cold-start data before RL. DeepSeek-R1 achieves performance comparable to OpenAI-o1 across math, code, and reasoning tasks.

發布日期
2025-01-20
參數規模
671.0B
上下文長度
164K
支援模態
text

能力雷達圖

80
general
50
coding
90
reasoning
60
science估算
78
agents
0
multimodal

缺少專門科學評測時,Science 由 LLM Stats 科學得分或推理能力估算。

排行榜排名

暫無排名資料

基準測試分數 (LLM Stats)

(LLM Stats (zeroeval))

Math

MATH-50095.9%自報
AIME 202486.7%自報

Reasoning

GPQANYU + Cohere + Anthropic (2023)73.3%自報
LiveCodeBench50.0%自報

AA 評測指數

(Artificial Analysis)

暫無 AA 評測資料

LLM Stats 分類評分

(LLM Stats (zeroeval))
Math
90
Reasoning
80
General
80
Physics
70
Biology
70
Chemistry
70
Code
50

定價

輸入價格$0.2574 / 1M tokens
輸出價格$1.0287 / 1M tokens
混合價格(3:1)$0.45023 / 1M tokens

速度

暫無速度資料

供應商價格排行

供應商價格排行

1 個供應商

供應商輸入輸出
1DeepSeek主要
$0.2574
$1.0287

比較該模型在不同 API 供應商之間的定價。

外部連結