GPT-5.4 Nano: token counter & pricing
OpenAI · exact (uses official tokenizer) · pricing as of 2026-05-31.
- Provider
- OpenAI
- API model ID
gpt-5.4-nano- Context window
- 400,000 tokens
- Input price
- $0.20 per 1M tokens
- Output price
- $1.25 per 1M tokens
- Tokenizer accuracy
- exact (uses official tokenizer)
- Pricing as of
- 2026-05-31
Open the counter to count tokens for GPT-5.4 Nano in real time.
What is GPT-5.4 Nano?
GPT-5.4 Nano is the cheapest tier in the GPT-5.4 family, designed for high-volume cost-sensitive workloads with the same o200k_base tokenizer and 400K context as the rest of the family. $0.20 input / $1.25 output per 1M tokens.
It sits between GPT-5 Nano ($0.05 input, cheapest in the catalog) and GPT-5 Mini ($0.25 input). For workloads where you want a small step up from base GPT-5 Nano's quality without committing to mini-tier pricing, GPT-5.4 Nano is the rational pick.
How tokens are counted here
OpenAI's o200k_base tokenizer. Browser-side via js-tiktoken. Exact.
Pricing notes
$0.20 input / $1.25 output per 1M. Cached input $0.02/M, effectively free at scale.
For 1,000 input + 200 output tokens: $0.000450 per call, $450 per 1M calls.
When to use GPT-5.4 Nano
- Classification, extraction, labeling where you want better reasoning than base GPT-5 Nano.
- First-pass filtering in agent pipelines that escalate to higher tiers on hard cases.
- Real-time UX where Nano-class latency matters.
When not to use it:
- Pure cost extremes. GPT-5 Nano at $0.05 input is 4× cheaper.
- Multi-step reasoning, escalate to GPT-5.4 Mini at minimum.
- Tasks where the older GPT-4.1 Nano ($0.10/$0.40) is sufficient and saves you 50% on input.
Common questions
GPT-5.4 Nano vs GPT-5 Nano vs GPT-4.1 Nano?
| Model | Input | Output | When |
|---|---|---|---|
| GPT-5 Nano | $0.05 | $0.40 | Cheapest, routine work |
| GPT-4.1 Nano | $0.10 | $0.40 | 1M context window |
| GPT-5.4 Nano | $0.20 | $1.25 | Best reasoning at "nano" tier |
Three Nano-tier models with three distinct positions. Pick by what your workload actually needs.
Is the cached-input discount real?
Yes, $0.02/M cached vs $0.20/M standard is a 10× discount. On agent loops with stable system prompts (most production agents), this is the biggest cost lever in the family.
What about the 35% Opus 4.8 tokenizer issue?
Doesn't apply here. All GPT-5 family models (including 5.4 Nano) use o200k_base which has stable, well-characterized tokenization across the family.
Compare GPT-5.4 Nano to other models
- GPT-5.5 (OpenAI, $5.00/$30.00)
- GPT-5.5 Pro (OpenAI, $30.00/$180.00)
- GPT-5.4 (OpenAI, $2.50/$15.00)
- GPT-5.4 Mini (OpenAI, $0.75/$4.50)
- GPT-5.4 Pro (OpenAI, $30.00/$180.00)
- GPT-5.3 (OpenAI, $1.75/$14.00)
- GPT-5.2 (OpenAI, $1.75/$14.00)
- GPT-5.2 Pro (OpenAI, $21.00/$168.00)
- GPT-5.1 (OpenAI, $1.25/$10.00)
- GPT-5 (OpenAI, $1.25/$10.00)
- GPT-5 Mini (OpenAI, $0.25/$2.00)
- GPT-5 Nano (OpenAI, $0.05/$0.40)
- GPT-5 Pro (OpenAI, $15.00/$120.00)
- GPT-4.1 (OpenAI, $2.00/$8.00)
- GPT-4.1 Mini (OpenAI, $0.40/$1.60)
- GPT-4.1 Nano (OpenAI, $0.10/$0.40)
- o3 (OpenAI, $2.00/$8.00)
- o3-mini (OpenAI, $1.10/$4.40)
- o3-pro (OpenAI, $20.00/$80.00)
- o4-mini (OpenAI, $1.10/$4.40)
- GPT-4o (OpenAI, $2.50/$10.00)
- GPT-4o mini (OpenAI, $0.15/$0.60)
- GPT-4 Turbo (OpenAI, $10.00/$30.00)
- Llama 3.1 8B (Meta, $0.18/$0.18)
- Gemini 3.1 Flash-Lite (Google, $0.25/$1.50)
- DeepSeek V3 (DeepSeek, $0.27/$1.10)