GPT Models

📖 5 min read openaigptmodelsreferencevendor-comparison Key Info

Deep comparison of GPT-5.5 vs GPT-5.4 vs GPT-5.4 mini vs GPT-5.4 nano — capabilities, pricing, context, reasoning levels, specialized models (Image-2, Realtime, Whisper, TTS), and model selection guide.

Key Takeaways

• Four current GPT tiers: GPT-5.5 ($5/$30, 1M ctx), GPT-5.4 ($2.50/$15, 1M), GPT-5.4 mini ($0.75/$4.50, 400K), GPT-5.4 nano ($0.20/$1.25, 400K)
• All GPT models support text + image input, tool use, streaming, structured outputs, and prompt caching
• GPT-5.5 has configurable reasoning levels (none/low/medium/high/xhigh) for balancing speed vs depth
• Specialized models: GPT Image 2 (image), Realtime-2 (voice), Sora (video), Whisper/TTS (speech)

On this page 11

OverviewCurrent GPT Models — May 2026Reasoning LevelsSpecialized ModelsGPT Image 2 — Image GenerationRealtime API — Voice & AudioSora — Video GenerationWhisper / TTSModel Selection GuideCost OptimizationComparing Across Models

Current GPT Models — May 2026

Feature	GPT-5.5	GPT-5.4	GPT-5.4 mini	GPT-5.4 nano
Description	Flagship — new class of intelligence	Affordable professional tier	Strong mini for coding & agents	Fastest, cheapest
Model ID	gpt-5.5	gpt-5.4	gpt-5.4-mini	gpt-5.4-nano
Input Pricing	$5 / 1M tokens	$2.50 / 1M tokens	$0.75 / 1M tokens	$0.20 / 1M tokens
Cached Input	$0.50 / 1M tokens	$0.25 / 1M tokens	$0.075 / 1M tokens	$0.02 / 1M tokens
Output Pricing	$30 / 1M tokens	$15 / 1M tokens	$4.50 / 1M tokens	$1.25 / 1M tokens
Context Window	1M tokens	1M tokens	400K tokens	400K tokens
Max Output	128K tokens	128K tokens	128K tokens	128K tokens
Reasoning Levels	none/low/medium/high/xhigh	none/low/medium/high/xhigh	none/low/medium/high/xhigh	none/low/medium
Vision (Image Input)	Yes	Yes	Yes	Yes
Tool Use	Functions, Web, File search, Computer use	Functions, Web, File search, Computer use	Functions, Web, File search, Computer use	Functions, Web
Streaming	Yes	Yes	Yes	Yes
Prompt Caching	Yes (10% of input)	Yes (10% of input)	Yes (10% of input)	Yes
Batch API (50% off)	Yes	Yes	Yes	Yes
Flex Processing	Yes	Yes	Yes	Yes
Knowledge Cutoff	Dec 1, 2025	Aug 31, 2025	Aug 31, 2025	Aug 31, 2025

Pricing History: GPT-5.5 dropped from $15/$75 (GPT-4 tier) to $5/$30. GPT-5.4 at $2.50/$15 offers near-flagship capability at a fraction of the price. All models have prompt caching at 10% of base input cost.

Reasoning Levels

GPT-5.5 and 5.4 models have configurable reasoning depth:

Level	Behavior	Cost	Latency	Best For
none	Standard response, no explicit reasoning	Lowest	Fastest	Simple Q&A, classification, routing
low	Light reasoning for moderate problems	Low	Fast	Code completion, summarization
medium	Balanced depth — good default	Medium	Medium	Analysis, code review, research
high	Deep reasoning for complex tasks	High	Slower	Architecture design, debugging
xhigh	Maximum reasoning — spends significant tokens “thinking”	Highest	Slowest	Hard math, complex multi-step problems

response = client.responses.create(
    model="gpt-5.5",
    input="Design a distributed rate limiter...",
    reasoning={"effort": "high"}  # Controls thinking depth
)

Specialized Models

GPT Image 2 — Image Generation

Feature	Detail
Model ID	gpt-image-2
Input (image)	$8 / 1M tokens ($2 cached)
Output (image)	$30 / 1M tokens
Input (text)	$5 / 1M tokens ($1.25 cached)
Use Cases	Product images, illustrations, design mockups, photo editing

Realtime API — Voice & Audio

Model	Use Case	Pricing
GPT Realtime 2	Voice agents, interactive audio	Audio: $32 in /$64 out per 1M. Text: $4 in /$24 out
GPT Realtime Translate	Live speech-to-speech translation	$0.034/min
GPT Realtime Whisper	Streaming speech-to-text	$0.017/min
GPT-4o Transcribe	High-quality speech-to-text	Pay-per-use
GPT-4o mini TTS	Text-to-speech generation	Pay-per-use

Sora — Video Generation

Cinematic video generation available via ChatGPT Pro and API. Pricing varies by resolution and duration.

Whisper / TTS

Traditional speech-to-text (Whisper) and text-to-speech (TTS) models available at lower cost than Realtime API variants.

Model Selection Guide

What matters most?
│
├─ Maximum quality, complex reasoning → GPT-5.5
│   Use when: R&D, architecture work, deep analysis
│   Cost: $5/$30 per 1M. Batch: $2.50/$15
│
├─ Best value for production → GPT-5.4
│   Use when: most APIs, coding, content, analysis
│   Cost: $2.50/$15 per 1M. Batch: $1.25/$7.50
│
├─ Cost-efficient at scale → GPT-5.4 mini
│   Use when: high-volume, computer use, subagents
│   Cost: $0.75/$4.50 per 1M. Batch: $0.375/$2.25
│
├─ Fastest, cheapest → GPT-5.4 nano
│   Use when: classification, routing, simple automation
│   Cost: $0.20/$1.25 per 1M
│
├─ Generate images → GPT Image 2
├─ Real-time voice/audio → Realtime API
├─ Speech-to-text → Realtime Whisper or GPT-4o Transcribe
└─ Text-to-speech → GPT-4o mini TTS

Cost Optimization

Strategy	Savings	When to Apply
Prompt Caching	90% on input	Repeated system prompts, same-context queries
Batch API	50% all token costs	Async, non-urgent workloads
Flex Processing	Lower cost	Non-production, lower-priority tasks
Model Routing	30-70%	Route to nano/mini for simple tasks, 5.5 for complex
Data Residency	+10% surcharge	Opt-in for regional processing compliance

Comparing Across Models

For a broader comparison across GPT, Claude, Gemini, and DeepSeek, see the Models Decision Guide.