Verified65% confidenceFactTime unknown

GPT-3 has a vocabulary of approximately 50,000 tokens.

3

Sources

65%

Confidence

Long-term

Relevance

6/3/2026

First Seen

Sources

Demystifying Transformer: A Word-Continuation Function, Deconstructed

bilibiliAI算法工程师Fiona

Related Entities

Related Claims

VerifiedGPT-3拥有1750亿参数80% similar UnverifiedGPT 系列模型的词表通常包含约 50,000–100,000 个 token79% similar UnverifiedGPT-2的BPE在字节级别操作，其merge表包含约5万条规则，最终词表为50,257个token78% similar

Cite This Claim

Stable URI

https://kongchang.com/claim/40177

API

curl https://kongchang.com/api/v1/knowledge/claims/40177

MCP

get_claim(id=40177)