GPT-5.4 Pricing Deep Dive: EvoLink 20% Discount, Cached Input & >272K Tier Rules
GPT-5.4 API Pricing: Per-Token Cost Breakdown for 2026
If you are planning API budget for GPT-5.4, this guide covers exact per-token pricing, GPT-5.x price history, cached input savings, and practical cost optimization strategies.
GPT-5.4 Pricing at a Glance
| Tier | Input (per 1M) | Output (per 1M) | Cached Input (per 1M) | Context |
|---|---|---|---|---|
| OpenAI direct | $2.50 | $15.00 | $0.25 | 1.05M / 128K max output |
| EvoLink (20% off) | $2.00 | $12.00 | $0.20 | 1.05M / 128K max output |
No monthly minimum. Pay per token, use one API key across all models.
GPT-5.x Pricing History
| Model | Release | Price (Input / Output, per 1M tokens) | Context | Notes |
|---|---|---|---|---|
| GPT-5.0 | Aug 2025 | $1.25 / $10.00 | 400K context / 128K max output | Launch pricing |
| GPT-5.1 | Nov 2025 | $1.25 / $10.00 | 400K | Same price, same core context tier |
| GPT-5.2 | Dec 2025 | $1.75 / $14.00 | 400K | 40% increase for stronger reasoning |
| GPT-5.2 Pro | Dec 2025 | $21.00 / $168.00 | 400K | Separately priced premium tier (Standard) |
GPT-5.3 (gpt-5.3-chat-latest / gpt-5.3-codex) | Mar 2026 | $1.75 / $14.00 | 400K | Listed API pricing (Standard) |
| GPT-5.4 | Mar 2026 | $2.50 / $15.00 | 1.05M context / 128K max output | Cached input: $0.25 / 1M |
Key pattern: OpenAI keeps base flagship tiers relatively stable across close generations (GPT-5.0 to GPT-5.1), but raises price on major reasoning upgrades (GPT-5.2) and prices dedicated premium tiers much higher (GPT-5.2 Pro).
How GPT-5.4 Pricing Compares to GPT-5.2 and GPT-5.1
GPT-5.4 is a price increase over GPT-5.2, justified by 1.05M context (vs 400K), stronger reasoning, and computer-use capabilities.
| GPT-5.1 | GPT-5.2 | GPT-5.4 | |
|---|---|---|---|
| Input | $1.25 | $1.75 | $2.50 |
| Output | $10.00 | $14.00 | $15.00 |
| Cached Input | $0.125 | $0.175 | $0.25 |
| Context | 400K | 400K | 1.05M |
| Max Output | 128K | 128K | 128K |
For teams already on GPT-5.2, the upgrade cost is modest on input (+43%) but small on output (+7%), while context jumps 2.6x.
Cached Input Pricing: Where the Real Savings Are
With 1.05M context, repeated system prompts can dominate input cost. Cached input pricing is where GPT-5.4 becomes practical for production:
| Tier | Standard Input | Cached Input | Savings |
|---|---|---|---|
| Base rate | $2.50 / 1M | $0.25 / 1M | 90% off |
| EvoLink discount | $2.00 / 1M | $0.20 / 1M | 90% off |
If your workload reuses large system contexts across requests, cached pricing can reduce effective input cost by 10x.
Competitive Price vs Capability Snapshot
Reference prices below are public list prices and can vary by tier and token-length bracket.
| Model | Price (Input / Output, per 1M tokens) | Context | Positioning |
|---|---|---|---|
| DeepSeek Chat | $0.27 / $1.10 (cache-miss input) | 64K | Budget and high-volume tasks |
| Gemini 2.5 Flash | $0.30 / $2.50 | 1M | Fast, low-cost, long-context tasks |
| GPT-5.1 | $1.25 / $10.00 | 400K | General-purpose prior generation |
| Gemini 3.1 Pro | $2.00-$4.00 / $12.00-$18.00 | 1M | Multimodal and long-context workloads |
| GPT-5.2 | $1.75 / $14.00 | 400K | Deep reasoning and coding |
| GPT-5.4 | $2.50 / $15.00 (cached input: $0.25) | 1.05M | Latest flagship with computer use |
| Claude Sonnet 4.6 | $3.00 / $15.00 | 1M (beta) | Coding and agentic tasks |
| Claude Opus 4.6 | $5.00 / $25.00 (base), $10.00 / $37.50 (>200K premium) | 1M (beta) | Research and complex reasoning |
GPT-5.4 output pricing ($15.00) is competitive with Gemini 3.1 Pro and significantly below Claude Opus 4.6. For teams routing through EvoLink, the discounted $12.00 output rate makes GPT-5.4 one of the best value flagship models available.
EvoLink GPT-5.4 Pricing: 20% Below Base Rate
EvoLink routes GPT-5.4 at a 20% discount from OpenAI base rate pricing:
| Base Rate | EvoLink Price | You Save | |
|---|---|---|---|
| Input | $2.50 / 1M | $2.00 / 1M | $0.50 / 1M |
| Output | $15.00 / 1M | $12.00 / 1M | $3.00 / 1M |
| Cached Input | $0.25 / 1M | $0.20 / 1M | $0.05 / 1M |
No subscription, no monthly minimum. One API key works across GPT-5.4, GPT-5.2, Claude, Gemini, and 200+ other models.
Cost Optimization Strategies
1. Use Prompt Caching Aggressively
With 1M+ context, repeated system prompts can dominate input cost. Keep stable context blocks identical across requests to maximize cached-token discounts.
2. Route by Task Complexity
Not every request needs extreme reasoning. Send simple requests to lower-cost models (GPT-5.1, DeepSeek Chat, Gemini Flash), and reserve GPT-5.4 for hard tasks.
3. Watch Token Efficiency
Larger context does not mean every task should consume larger context. Measure whether 1M context improves success rate enough to justify higher input spend.
4. Optimize for Cost per Task, Not Cost per Token
A higher-priced model that solves in one pass can still be cheaper than a low-cost model requiring retries. Track total cost per successful outcome.
FAQ
How much does GPT-5.4 API cost per call?
A typical API call with 2,000 input tokens and 500 output tokens costs about $0.0125 at base rate, or $0.0100 through EvoLink. With cached input, the same call drops to around $0.008.
How much does GPT-5.4 cost compared to GPT-5.2?
GPT-5.4 input is 43% more expensive ($2.50 vs $1.75 per 1M), but output is only 7% higher ($15.00 vs $14.00). You get 2.6x the context window (1.05M vs 400K). Through EvoLink, the gap narrows further: $2.00/$12.00 vs $1.75/$14.00 — meaning EvoLink GPT-5.4 output is actually cheaper than base-rate GPT-5.2 output.
Is GPT-5.4 cheaper than Claude Opus 4.6?
Yes. GPT-5.4 is $2.50/$15.00 vs Claude Opus 4.6 at $5.00/$25.00 (base) or $10.00/$37.50 (>200K premium). Through EvoLink, GPT-5.4 at $2.00/$12.00 is less than half the cost of Opus 4.6.
What is GPT-5.4 cached input pricing?
Cached input tokens cost $0.25 per 1M at base rate, a 90% discount from standard input. Through EvoLink, cached input is $0.20 per 1M. This matters for workloads with large, repeated system prompts.
How does EvoLink GPT-5.4 pricing work?
EvoLink charges per token with no monthly minimum and no subscription. Buy credits, get one API key, and use it across GPT-5.4, GPT-5.2, Claude, Gemini, and 200+ other models. EvoLink is 100% compatible with the OpenAI SDK — just change the base URL.
Is GPT-5.4 worth upgrading from GPT-5.2?
If you need 1M+ context, computer-use capabilities, or stronger reasoning, the upgrade is justified — especially through EvoLink where the output price ($12.00) is actually lower than GPT-5.2 base rate ($14.00). If 400K context is sufficient for your workload, GPT-5.2 remains a strong value option.
