TokenMix Research Lab · 2026-04-03
GPT-5 API Pricing 2026: 5.5, 5.4, Mini Costs, Batch Math
Last Updated: 2026-04-30
Author: TokenMix Research Lab
Data checked: 2026-04-30
GPT-5 API pricing now has three public flagship routes: GPT-5.5 at $5/$30, GPT-5.4 at $2.50/$15, and GPT-5.4 mini at $0.75/$4.50 per 1M tokens. Cache reads are 90% cheaper than base input, and Batch API cuts input and output by 50%.
According to OpenAI's official API pricing page, GPT-5.5 costs $5 input, $0.50 cached input, and $30 output per 1M tokens. GPT-5.4 costs $2.50 input, $0.25 cached input, and $15 output. GPT-5.4 mini costs $0.75 input, $0.075 cached input, and $4.50 output. OpenAI also states that standard rates shown apply to context lengths under 270K, Batch API saves 50% on inputs and outputs, and data residency adds 10%.
Table of Contents
- Quick Answer
- Confirmed GPT-5 Pricing Facts
- GPT-5 Price Table
- Standard vs Cached vs Batch
- GPT-5.5 vs GPT-5.4 vs GPT-5.4 Mini
- Cost per Task
- Monthly Cost Scenarios
- When To Escalate
- GPT-5 vs Claude, Gemini, DeepSeek
- When TokenMix.ai Fits
- Final Recommendation
- FAQ
- Related Articles
- Sources
Quick Answer
| Question | Answer |
|---|---|
| Cheapest GPT-5 API route | GPT-5.4 mini at $0.75 input and $4.50 output per 1M tokens. |
| Default GPT-5 route | GPT-5.4 at $2.50 input and $15 output. |
| Premium GPT-5 route | GPT-5.5 at $5 input and $30 output. |
| Cached input discount | 90% off base input for all three listed GPT-5 routes. |
| Batch discount | 50% off input and output for async Batch API jobs. |
| Key context caveat | The public pricing page says listed standard rates apply under 270K context length. |
Confirmed GPT-5 Pricing Facts
| Claim | Status | Practical meaning | Source |
|---|---|---|---|
| GPT-5.5 is $5/$30 per 1M tokens | Confirmed | It is the premium OpenAI route. | OpenAI pricing |
| GPT-5.4 is $2.50/$15 per 1M tokens | Confirmed | It is the balanced OpenAI default. | OpenAI pricing |
| GPT-5.4 mini is $0.75/$4.50 per 1M tokens | Confirmed | It is the budget GPT-5 route. | OpenAI pricing |
| Cached input is 10% of base input | Confirmed by listed cached-input prices | Repeated context can change the cost curve. | OpenAI pricing |
| Batch API saves 50% on inputs and outputs | Confirmed | Best for offline evaluation, summarization, and enrichment. | OpenAI pricing |
| Data residency adds 10% | Confirmed pricing toggle | Regulated deployments need this in cost forecasts. | OpenAI pricing |
| Old four-tier GPT price tables should be used | Not supported by current public page checked | Do not model current GPT-5 spend from stale page copies. | TokenMix analysis |
GPT-5 Price Table
All prices are per 1M tokens in USD.
| Model | Input | Cached input | Output | Best use |
|---|---|---|---|---|
| GPT-5.4 mini | $0.75 | $0.075 | $4.50 | Cheap OpenAI route, agents, bulk simple tasks |
| GPT-5.4 | $2.50 | $0.25 | $15.00 | Default production route |
| GPT-5.5 | $5.00 | $0.50 | $30.00 | Hard coding, professional work, escalation |
This is the clean current picture. Older pages that mention a four-tier GPT-5.4 lineup are not aligned with the public OpenAI pricing page checked on 2026-04-30.
Standard vs Cached vs Batch
OpenAI cost levers are simple but powerful.
| Model | Standard input | Cached input | Batch input | Standard output | Batch output |
|---|---|---|---|---|---|
| GPT-5.4 mini | $0.75 | $0.075 | $0.375 | $4.50 | $2.25 |
| GPT-5.4 | $2.50 | $0.25 | $1.25 | $15.00 | $7.50 |
| GPT-5.5 | $5.00 | $0.50 | $2.50 | $30.00 | $15.00 |
Effective input with repeated context:
| Model | Base input | 70% cached input | 90% cached input |
|---|---|---|---|
| GPT-5.4 mini | $0.75 | $0.2775 | $0.1425 |
| GPT-5.4 | $2.50 | $0.9250 | $0.4750 |
| GPT-5.5 | $5.00 | $1.8500 | $0.9500 |
Formula:
effective input = uncached_share * base_input + cached_share * cached_input
The catch: cached input only helps when the prefix actually repeats. It does not reduce output tokens.
GPT-5.5 vs GPT-5.4 vs GPT-5.4 Mini
| Decision factor | GPT-5.4 mini | GPT-5.4 | GPT-5.5 |
|---|---|---|---|
| Price tier | Budget | Balanced | Premium |
| Input vs GPT-5.4 | 70% cheaper | Baseline | 2x higher |
| Output vs GPT-5.4 | 70% cheaper | Baseline | 2x higher |
| Best default role | First-pass work | Main production route | Escalation route |
| Avoid when | Quality failures cost more than token savings | You need the highest quality | Request is simple or high-volume |
The practical routing policy:
| Request type | First route | Escalate if |
|---|---|---|
| Classification | GPT-5.4 mini | The label is high-risk or ambiguous. |
| Extraction | GPT-5.4 mini | The document is messy or legally sensitive. |
| User-facing answer | GPT-5.4 | The answer must be near-perfect. |
| Coding task | GPT-5.4 | Multi-file reasoning fails. |
| Deep professional analysis | GPT-5.5 | Start here only when failure cost is high. |
Cost per Task
These examples use standard non-batch pricing and no cache unless stated.
| Task | Token shape | GPT-5.4 mini | GPT-5.4 | GPT-5.5 |
|---|---|---|---|---|
| Simple chat reply | 500 input / 200 output | $0.001275 | $0.004250 | $0.008500 |
| Support answer | 2K input / 800 output | $0.005100 | $0.017000 | $0.034000 |
| RAG answer | 8K input / 500 output | $0.008250 | $0.027500 | $0.055000 |
| Code review | 20K input / 3K output | $0.028500 | $0.095000 | $0.190000 |
| Long cached agent turn | 100K input, 70% cached / 2K output | $0.036750 | $0.122500 | $0.245000 |
Batch version of the same code review:
| Model | Standard cost | Batch cost | Savings |
|---|---|---|---|
| GPT-5.4 mini | $0.028500 | $0.014250 | 50% |
| GPT-5.4 | $0.095000 | $0.047500 | 50% |
| GPT-5.5 | $0.190000 | $0.095000 | 50% |
Monthly Cost Scenarios
Assume 100M input tokens and 30M output tokens per month.
| Model | No cache | 70% input cache | Batch only | Batch plus 70% cache estimate |
|---|---|---|---|---|
| GPT-5.4 mini | $210.00 | $162.75 | $105.00 | $81.38 |
| GPT-5.4 | $700.00 | $542.50 | $350.00 | $271.25 |
| GPT-5.5 | $1,400.00 | $1,085.00 | $700.00 | $542.50 |
At this volume, GPT-5.5 costs 6.7x GPT-5.4 mini before cache and batch. Use it where quality changes the outcome, not as a default bulk model.
When To Escalate
| Situation | Stay on mini | Use GPT-5.4 | Escalate to GPT-5.5 |
|---|---|---|---|
| Bulk labels | Yes | Rarely | No |
| Support drafts | Often | Yes for final answer | Rarely |
| Code generation | For simple edits | Default | Hard multi-step tasks |
| Legal/finance summaries | No | Usually | High-stakes or ambiguous |
| Agent planning | Cheap first pass | Main route | Failed or high-value turns |
The key metric is cost per successful workflow, not cost per token.
GPT-5 vs Claude, Gemini, DeepSeek
| Model | Input | Cached input | Output | Main role |
|---|---|---|---|---|
| GPT-5.4 mini | $0.75 | $0.075 | $4.50 | Budget OpenAI route |
| GPT-5.4 | $2.50 | $0.25 | $15.00 | OpenAI default |
| GPT-5.5 | $5.00 | $0.50 | $30.00 | Premium OpenAI route |
| Claude Sonnet 4.6 | $3.00 | $0.30 | $15.00 | Balanced Claude route |
| Claude Opus 4.7 | $5.00 | $0.50 | $25.00 | Premium Claude route |
| Gemini 3.1 Pro | $2.00 | $0.20 | $12.00 | Premium Gemini route under 200K |
| DeepSeek V4 Flash | $0.14 miss | $0.0028 hit | $0.28 | Lowest-cost text route |
GPT-5.4 is competitive on quality and ecosystem fit. It is not the lowest-cost route. DeepSeek dominates raw token price, Gemini is strong under 200K, and Claude remains a strong quality benchmark.
When TokenMix.ai Fits
TokenMix.ai fits when GPT-5 is one route in a multi-model system. Instead of wiring separate clients for OpenAI, Claude, Gemini, DeepSeek, and other providers, you can use one OpenAI-compatible access layer and route by task.
| Need | Direct OpenAI | TokenMix.ai unified API |
|---|---|---|
| Native OpenAI feature coverage | Best path | Use when compatible |
| GPT plus Claude/Gemini/DeepSeek | Multiple integrations | One routing layer |
| Cost-aware fallback | Build yourself | Centralized policy |
| Payment flexibility | OpenAI billing path | Useful when direct billing is hard |
| Price comparison | Manual | Compare model families |
Use the AI API pricing hub for cross-provider price decisions.
Final Recommendation
Use GPT-5.4 mini for cheap first-pass work, GPT-5.4 as the default OpenAI production route, and GPT-5.5 only for hard tasks where a better answer is worth 2x the GPT-5.4 token price.
FAQ
How much does GPT-5.5 API cost?
GPT-5.5 costs $5 input, $0.50 cached input, and $30 output per 1M tokens on OpenAI's public pricing page checked on 2026-04-30.
How much does GPT-5.4 API cost?
GPT-5.4 costs $2.50 input, $0.25 cached input, and $15 output per 1M tokens. It is the balanced default in the current GPT-5 API lineup.
How much does GPT-5.4 mini cost?
GPT-5.4 mini costs $0.75 input, $0.075 cached input, and $4.50 output per 1M tokens. It is the cheapest GPT-5 route listed on the public pricing page.
Does GPT-5 API cached input save 90%?
Yes. The cached input prices shown by OpenAI are 10% of base input prices for GPT-5.5, GPT-5.4, and GPT-5.4 mini.
Does Batch API reduce GPT-5 costs?
Yes. OpenAI states that Batch API saves 50% on inputs and outputs. Use it for async jobs where a user is not waiting.
Is GPT-5.5 worth the extra cost?
Use GPT-5.5 when failure is expensive: hard coding, professional analysis, and multi-step problems. For routine tasks, GPT-5.4 or GPT-5.4 mini is usually more cost-efficient.
Is GPT-5 cheaper than Claude?
GPT-5.4 mini is cheaper than current Claude 4.5/4.6 routes. GPT-5.4 is cheaper than Claude Sonnet 4.6 on input but matches it on output. GPT-5.5 is more expensive than Claude Opus 4.7 on output.
Should I use GPT-5 directly or through a gateway?
Use OpenAI directly for native feature coverage and direct billing. Use a gateway such as TokenMix.ai when your app needs GPT plus Claude, Gemini, DeepSeek, routing, fallback, and centralized cost tracking.
Related Articles
- AI API Pricing 2026: 16 Models, Cache, Batch, Routing Hub
- OpenAI API Pricing 2026: GPT-5.5, Realtime, Image Costs
- Claude API Pricing 2026: Opus, Sonnet, Haiku Costs Compared
- Google Gemini API Pricing 2026: 3.1 Pro, Flash, Batch Costs
- DeepSeek API Pricing 2026: V4 Costs, Cache Hits, R1 Changes
- OpenAI-Compatible API Gateway: 9 Providers, One SDK Guide
- AI API Gateway 2026: 7 LLM Routing and Fallback Options
- OpenAI API No Credit Card 2026: 5 Legal Ways To Get Access