ChatGPT / OpenAI — Cheat Sheet 2026

✦Current OpenAI Models — May 2026

GPT-5.5 Flagship

gpt-5.5 · gpt-5.5-pro

Most capable. Complex reasoning and coding. Use for highest-quality outputs and agentic workflows.

Vision + toolsPrompt cache 90% offPriority tier available

GPT-5.4 Series Balanced

gpt-5.4 · gpt-5.4-mini · gpt-5.4-nano · gpt-5.4-pro

Professional work and coding. Adaptive reasoning, improved latency. Mini/nano for cost-sensitive apps.

GPT-5.4-mini: lower latencyGPT-5.4-nano: cheapest

GPT-4.1 Series Best for code

gpt-4.1 · gpt-4.1-mini · gpt-4.1-nano

Significantly better at coding than GPT-4o. GPT-4.1: $2/$8/M · Mini: $0.40/$1.60/M · Nano: cheap.

1M context33K output (mini)Prompt cache 75% off

GPT-4o / 4o-mini Proven

gpt-4o · gpt-4o-mini

Reliable, widely integrated. 4o-mini: $0.15/$0.60/M. Prompt cache 50% off. Vision + audio.

128K contextMultimodal

o3 Reasoning

Frontier reasoning. 80% price cut — now $0.40/$1.60/M. Best for math, science, coding, vision tasks.

200K contextReasoning tokens billed

o4-mini Fast Reasoning

o4-mini

Compact, efficient reasoning. Excels at math, coding, visual tasks. Great price/performance ratio.

200K contextFine-tunable$1.10/$4.40/M

💡Start with gpt-5.5 for complex tasks, gpt-5.4-mini or gpt-4.1-nano for cost-sensitive / high-volume

🧠o-series reasoning tokens: billed as output but not returned. A 500-token response may use 2000+ tokens. Monitor carefully.

💰Prompt caching: GPT-5 family 90% off · GPT-4.1 family 75% off · GPT-4o/o-series 50% off. Cache persists 5–10 min.

⚡Quickstart Code

Python (openai SDK)

pip install openai from openai import OpenAI client = OpenAI(api_key="sk-...") resp = client.chat.completions.create( model="gpt-5.5", messages=[{"role":"user","content":"Hi"}] ) print(resp.choices[0].message.content)

JavaScript / Node

npm install openai import OpenAI from "openai" const client = new OpenAI({apiKey:"sk-..."}) const resp = await client.chat.completions.create({ model:"gpt-5.5", messages:[{role:"user",content:"Hi"}] })

cURL

curl api.openai.com/v1/chat/completions \ -H "Authorization: Bearer $OPENAI_API_KEY" \ -H "Content-Type: application/json" \ -d '{"model":"gpt-5.5","messages":[{"role":"user","content":"Hi"}]}'

📡API Endpoints

Base URL: api.openai.com/v1/

POST

/chat/completions

Main chat endpoint

POST

/responses

Responses API (newer, recommended)

POST

/assistants · /threads · /runs

Assistants API

POST

/images/generations

DALL-E / gpt-image-1

POST

/audio/speech · /audio/transcriptions

TTS + Whisper STT

POST

/embeddings

Text embeddings

POST

/fine_tuning/jobs

Fine-tuning (GPT-4.1, 4o, o4-mini)

POST

/batches

Async batch — 50% off, 24h SLA

POST

/moderations

Content moderation check

GET

/models

List available models

POST

/vector_stores

File search vector stores

⚙️Key Parameters

Chat Completions

modelstring — required

messagesarray — [{role, content}]

temperature0–2 · 0=deterministic, 1=balanced

streambool — SSE streaming

response_format{type:"json_schema"} structured output

toolsarray — function definitions

tool_choice"auto" | "none" | force function

parallel_tool_callsbool — multiple tools at once

seedint — deterministic outputs (best effort)

reasoning_effort"low"|"medium"|"high" — o-series only

max_completion_tokensint — max response tokens

logprobsbool — token log probabilities

userstring — unique user ID for abuse detection

Structured Output

📐Use json_schema with strict:true for guaranteed JSON. Avoids hallucinated fields.

💎ChatGPT Plans / Tiers

Free $0/mo

GPT-4o limitedLimited searchesBasic voice

Plus $20/mo

GPT-5.5 accesso3/o4-miniCanvasImage genAdvanced voiceSearch

Pro $200/mo

Unlimited GPT-5.5o3 extendedPriority accessAll features

Team $25/user/mo

All Plus featuresAdmin consoleData privacyHigher limits

Enterprise Custom

SSO/SAMLSOC 2 Type IICustom retentionDedicated support

API Rate Tiers

Tier 1 ($5 spent): 500 RPM · 30K TPM
Tier 2 ($50): 5K RPM · 450K TPM
Tier 3 ($100): 5K RPM · 800K TPM
Tier 4 ($250): 10K RPM · 2M TPM
Tier 5 ($1K): 30K RPM · 150M TPM

💬ChatGPT Web Features

🧠

Memory

Remembers info across conversations

Plus+

🎨

Canvas

Collaborative writing and code editing

Plus+

🤖

GPTs / Custom GPTs

Build and share custom assistants

All

📁

Projects

Organise chats, share files + instructions

All

🔍

ChatGPT Search

Real-time web search with citations

Plus+

🖼️

Image Generation

DALL-E 3 + gpt-image-1 in chat

Plus+

🎤

Advanced Voice Mode

Real-time voice with emotion detection

Plus+

💻

Code Interpreter

Python sandbox, charts, data analysis

Plus+

📄

File Uploads

PDFs, images, code, spreadsheets

All

🤖Assistants API

🧵Persistent AI with threads, tools, and file access. Beta — using client.beta.assistants

AssistantPersistent AI with instructions, model, tools

ThreadConversation history — add messages, run assistant

RunExecution on a thread — poll for completion

Code Interp.Run Python, create files, process data

File SearchVector store semantic search over files (RAG)

Vector StoreStore + search file embeddings for File Search

Quick Create

a = client.beta.assistants.create( model="gpt-5.5", instructions="You are a helpful analyst", tools=[{"type":"code_interpreter"}] ) t = client.beta.threads.create()

Audio APIs

🎵Whisper STT: gpt-4o-transcribe — audio → text, multi-language

🔊TTS: tts-1, tts-1-hd, gpt-4o-mini-tts — 6 voices, MP3/Opus

⚡Realtime API: gpt-4o-realtime-preview — bidirectional audio WebSocket, sub-300ms

🔧Function Calling, Structured Output & Prompting Tips

Function Calling

tools = [{ "type": "function", "function": { "name": "get_weather", "description": "Get weather", "parameters": { "type": "object", "properties": { "city": {"type":"string"}}, "required": ["city"]}}}] # Check resp.choices[0].message.tool_calls

Structured JSON Output

resp = client.chat.completions.create( model="gpt-5.5", messages=messages, response_format={ "type": "json_schema", "json_schema": { "name": "result", "schema": {...your schema...}, "strict": True}})

Keyboard Shortcuts

Ctrl+Shift+O — New conversation
Ctrl+Shift+S — Toggle sidebar
Ctrl+K — Search conversations
Ctrl+Shift+C — Copy last code block
Shift+Enter — New line (not send)
Esc — Stop generating

Prompting Tips for GPT / o-series

🎯System prompt first. Put all persistent instructions in the system role — highest priority, saves tokens.

📋Few-shot examples. 2–5 input/output pairs. GPT-5 follows patterns very quickly.

🌡️Temperature. 0 for extraction/classification. 0.7–1 for creative writing. Above 1 degrades quality.

📐Structured output. Always use json_schema with strict:true. Eliminates parsing errors.

🧠o-series. Don't add chain-of-thought — model reasons internally. Use reasoning_effort:"high" for hard problems.

💾Prompt caching. Keep static content at the start of your prompt for best cache hit rates.

Troubleshooting

🔑401 Unauthorized — Invalid API key. Check OPENAI_API_KEY, no trailing whitespace.

⏱️429 Rate Limit — Implement exponential backoff. Consider Batch API (50% off, async).

⏰Timeout — Use stream:true for long outputs. Increase client timeout.

ChatGPT & OpenAI API