LANGUAGE MODEL OpenAI

GPT-3

Language Models are Few-Shot Learners

A language model that became dramatically more capable than its predecessors simply by being roughly 100 times larger and trained on roughly 100 times more text. It demonstrated that you could often get a model to do a new task by just describing the task in plain English and showing a few examples — no retraining required.

Intelligence
Below avg
Speed
Medium
107 tok/s output
Cost
Moderate
$0.50 in / $1.50 out
Context
2K
Up to 2,048 tokens
How are Intelligence, Speed & Cost bucketed?
Intelligence and Speed buckets are percentile ranks on Artificial Analysis. Cost buckets are fixed dollar thresholds keyed off output-token price ($/M out).
Intelligence
  • Top 1%≤ 1%
  • Top 5%≤ 5%
  • Top 10%≤ 10%
  • Good≤ 25%
  • Medium≤ 50%
  • Below avg> 50%
Speed
  • Top 1%≥ 345 tok/s
  • Top 5%≥ 237 tok/s
  • Top 10%≥ 196 tok/s
  • Good≥ 146 tok/s
  • Medium≥ 90 tok/s
  • Slow< 90 tok/s
Cost
  • Freeopen weights · self-host
  • Low< $1 / M out
  • Moderate$1–5 / M out
  • High≥ $5 / M out

Why it matters

GPT-3 is the moment "large language model" became a coherent product category. Every API-priced LLM business — OpenAI, Anthropic, Cohere, Mistral, the API arms of Google and Meta — uses GPT-3's pricing model (per-token), interface model (text-in, text-out), and capability framing (few-shot prompting). Without this paper, the 2022–2026 AI investment cycle does not happen on the same timeline.

Core Capabilities

Long Documents
Handles entire codebases, books, and multi-doc RAG.
Generative
Produces images, video, audio, or other media.

Context Window

2k tokens
short prompt
4k Chat 聊天
32k Long docs 长文档
128k Books 整本书
400k Multi-doc 多文档
1M Codebase 整个代码库
10M
2k

Availability

API
Available
Product / App
Not available
Open Source
Not released
Enterprise
Contact sales

Pricing Model

Pay per token
Input and output billed separately.
Pay-per-token

Capability / Performance

Where this model sits relative to the middle 60% of models in the tree. All scores are 0–10 (higher is better).

Lower 20% Upper 80% This model
Reasoning
AA Intelligence Index · scaled to 10
1.7
5.6
1.3
Coding
AA Coding Index · scaled to 10
1.8
4.3
1.5
Context / memory
Context window size · log-scaled
6.0
9.0
0.0
Cost efficiency
Input price ($/M tokens) · cheaper scores higher
6.2
10.0
10.0
Lower 20% 20th percentile — 20% of models score below this This model Where the current model lands Upper 80% 80th percentile — only 20% of models score above this Percentile boundaries are computed across every model in the tree that reports the underlying benchmark for each capability.

What it feels like

  • 175B parameters — 100x larger than GPT-2; the moment scale-as-progress became conventional wisdom
  • Few-shot in-context learning emerged as a property of scale — pattern that reframed the whole field
  • Released as the paid OpenAI API in mid-2020 — first commercial frontier-tier LLM
  • Inspired the Codex spinoff, then GitHub Copilot, then the entire AI-coding-tools wave
  • Foundation for InstructGPT (RLHF) and ChatGPT — the productisation that broke out of research
  • By 2025 measurements, GPT-3 trails on every benchmark — historical interest only

Best use cases

  • Reading the seminal scaling paper to understand modern LLM emergence
  • Few-shot prompting research — GPT-3 popularised the technique
  • Citation in any work on scaling laws, in-context learning, or LLM history

Tools to try

Not ideal for

  • Production deployment in 2025 — every successor is cheaper and better
  • Reasoning, coding, or tool-use workloads — modern Sonnet/Haiku/GPT-4o dominate

Model Evolution

View full evolution tree →

Brown, T. B. · Mann, B. · Ryder, N. · Subbiah, M. · et al.