GPT-5.4 nano API Pricing
Deprecated: OpenAI announced on October 1, 2026 that gpt-5.4-nano leaves the API on April 1, 2027, and names GPT-6 Luna as the replacement. OpenAI's cheapest GPT-5.4-class model for simple high-volume tasks: $0.20/M input, $1.25/M output, and $0.02/M cached input. It is built for classification, extraction, ranking, routing, and small subagents where cost matters more than maximum capability.
GPT-5.4 nano API cost calculator.
Live calculator pre-loaded with GPT-5.4 nano rates, which apply until OpenAI shuts the model down on April 1, 2027. Use it for high-volume routing, extraction and classification, and to compare the bill against the replacement.
GPT-5.4 nano cost on real workloads.
Lead scoring classifier
Invoice field extraction
Support ticket routing
Bulk doc summary
Codebase triage lint pass
Counted with the o200k_base BPE encoding — the same tokenizer the API uses — entirely in your browser. The encoder loads on your first keystroke (~1 MB, one time); your text never leaves this page.
| Model | Input /M | Output /M | Effective blended | Context | Best for |
|---|---|---|---|---|---|
| GPT-5.4 nano Current | $0.20 cache $0.02 | $1.25 | $0.148 agentic 92/8 | 400K | Classification and extraction |
| GPT-5.4 mini | $0.75 cache $0.075 | $4.50 | $0.541 +3.6× | 400K | Subagents and computer-use tasks |
| GPT-5.4 | $2.50 cache $0.25 | $15.00 | $1.80 +12× | 1.05M | Affordable OpenAI frontier work |
| GPT-5.3-Codex | $1.75 cache $0.175 | $14.00 | $1.54 deprecated · +10× | 400K | Agentic coding tasks |
| Gemini 3.8 Flash | $0.75 cache $0.075 | $3.75 | $0.481 +3.2× · promo | 1M | Large multimodal context |
| DeepSeek V4 Pro | $1.32 cache $0.044 | $3.96 | $0.569 +3.8× | 1M | Low-cost reasoning workloads |
| DeepSeek V4.1 Flash | $0.30 cache $0.006 | $1.20 | $0.15 same | 1M | Bulk low-cost traffic |
GPT-5.4 nano has published scores on 2 benchmarks we track; its best placing is #69 of 105 on ARC-AGI-2. All of these scores come from independent leaderboards, not from the lab.
| Benchmark | Score | Rank | Leader | Source |
|---|---|---|---|---|
| ARC-AGI-2 Reasoning | 5.7% ARC Prize verified, best result | #69 of 105 | GPT-6 Astra · 95% | Official leaderboard ARC Prize Foundation |
| ARC-AGI-1 Reasoning | 51.5% ARC Prize verified, best result | #69 of 99 | 6 tied · 98.5% | Official leaderboard ARC Prize Foundation |
GPT-5.4 nano API pricing FAQ.
Practical GPT-5.4 nano pricing questions, with OpenAI's published rates separated from workload assumptions.
Q · 01 When does GPT-5.4 nano stop working? +
gpt-5.4-nano and its snapshot gpt-5.4-nano-2026-03-17 on October 1, 2026 and recommends GPT-6 Luna, which lists at $0.10/M input and $0.50/M output against $0.20 and $1.25 here. Until the shutdown nano bills at the rates on this page.Q · 02 What is the standard GPT-5.4 nano API price? +
gpt-5.4-nano at $0.20/M input tokens, $0.02/M cached input tokens, and $1.25/M output tokens. This page stores the public list price in USD and marks the source as OpenAI's pricing docs.Q · 03 How much cheaper is GPT-5.4 nano than GPT-5.4 mini? +
$0.15/M while GPT-5.4 mini is about $0.54/M. That makes nano roughly 3.6x cheaper for the same token mix.Q · 04 How much cheaper are Batch and Flex? +
gpt-5.4-nano at half the standard rate: $0.10/M input, $0.01/M cached input, and $0.625/M output. That is the best fit for latency-tolerant bulk classification or extraction.Q · 05 Is Priority pricing available for GPT-5.4 nano? +
gpt-5.4-nano on OpenAI's pricing page. The Priority table lists gpt-5.5, gpt-5.4, and gpt-5.4-mini, so this page does not publish a Priority rate for nano.Q · 06 Does GPT-5.4 nano have regional pricing? +
10% uplift for gpt-5.4-nano and related GPT-5.4/5.5 models. Standard global routing uses the public standard rate.Q · 07 Does GPT-5.4 nano support computer use? +
Not supported for gpt-5.4-nano. Use GPT-5.4 mini or GPT-5.4 when browser or desktop automation is part of the workload.Q · 08 What release date is used here? +
gpt-5.4-nano-2026-03-17. This page uses 2026-03-17 as the model release date for pricing-history purposes.Q · 09 How accurate is the tokenizer estimate? +
4.875 characters per token for English text and labels the tokenizer as tiktoken-o200k_base. It is good for budget planning, but exact billing can differ by language, whitespace, code, tool calls, and image inputs.