OpenAI model API pricing
OpenAI gpt-5.4 API pricing
Direct answer
As of June 27, 2026, OpenAI's gpt-5.4 lists at $2.50 per 1M input tokens and $15.00 per 1M output tokens in the ByteCosts pricing index (A+ confidence, source last checked 2026-06-27), with no context window stated in this record, which is a planning estimate you should verify against the provider source before billing, because output volume, retries, and prompt-cache hit rate move the real gpt-5.4 bill far more than the headline input rate.
Estimate gpt-5.4 monthly cost - Open the calculator with your traffic →
gpt-5.4 price card
List prices per million tokens from the ByteCosts pricing index, graded A+ confidence. Output tokens are the expensive side of almost every bill.
| Price field | Value |
|---|---|
| Input | $2.50 / 1M tokens |
| Output | $15.00 / 1M tokens |
| Cached input (read) | $0.250 / 1M tokens |
| Cache write | not published |
| Cache write (1h) | not published |
| Context window | not in this record |
Context window and output limits
What gpt-5.4 can hold in a single request, and where ByteCosts stops short of a source-backed claim.
- Context window: not part of this committed pricing record; check the provider page for the current limit.
- Maximum output tokens: not a field in this committed pricing record. ByteCosts only states source-backed numbers, so treat the provider documentation as authoritative for the per-response output cap.
What it costs at real workloads
Each row is computed from this record's committed input and output token prices with the ByteCosts cost engine. Assumptions are shown inline.
| Workload | Assumed monthly tokens | Estimated token cost |
|---|---|---|
| Light reference | 1M input + 250K output tokens/month | $6.25 / month |
| Production reference | 10M input + 2.5M output tokens/month | $62.50 / month |
| Heavy reference | 100M input + 25M output tokens/month | $625 / month |
Best-fit workloads for gpt-5.4
Where this price profile is cost-effective, and where it is not, derived from the committed input, output, cache, and context fields.
- Good fit: Repeated-prefix agents and chat with a stable system prompt: cached input reads list at $0.250 per 1M, well below the $2.50 fresh-input rate.
- Good fit: Input-heavy, short-output work (RAG, extraction, classification, reranking pre-filters): output costs about 6.0x input here, so keeping responses tight protects the budget.
- Not ideal: High-volume long-form generation: output is about 6.0x the input price ($15.00 vs $2.50 per 1M), so verbose or reasoning-heavy responses dominate the invoice.
gpt-5.4 pricing across providers
gpt-5.4 is listed by 2 providers in the ByteCosts index. Per-million-token list prices, lowest input first.
| Provider | Input / 1M | Output / 1M |
|---|---|---|
| OpenAI (this page) | $2.50 | $15.00 |
| Vivgrid | $2.50 | $15.00 |
When gpt-5.4 is cost-effective
gpt-5.4 ranks 5 of 9 OpenAI records by combined list price ($17.50 for 1M input plus 1M output tokens). That puts it in the middle third of this provider's committed records.
Cached input reads are listed at $0.250 per 1M tokens, 10% of the fresh input rate, so repeated cached prompts have a lower committed input-token cost than fresh prompts.
No context window is committed for this record, so there is no long-context cost signal to apply.
Cheaper alternatives
Records with a lower combined input plus output list price than this one. Same-provider and same-model cross-provider records are preferred where available.
| Record | Input / 1M | Output / 1M | Combined / 1M + 1M | Detail page |
|---|---|---|---|---|
| OpenAI gpt-5.4-nano | $0.200 | $1.25 | $1.45 | /tools/ai-provider-pricing/openai--gpt-5.4-nano |
| OpenAI gpt-5.4-mini | $0.750 | $4.50 | $5.25 | /tools/ai-provider-pricing/openai--gpt-5.4-mini |
| OpenAI gpt-4o-transcribe | $2.50 | $10.00 | $12.50 | /tools/ai-provider-pricing/openai--gpt-4o-transcribe |
Evidence, freshness, and limitations
ByteCosts grades this OpenAI gpt-5.4 row A+: Read from the provider's official pricing page. The record was first tracked on 2026-05-30 and last checked on 2026-06-27, and the source URL stays attached so you can verify the provider source before a purchasing or architecture decision.
Every number here is a planning estimate, not a quote. Real invoices can differ because taxes, negotiated and volume discounts, minimum commitments, batch and priority tiers, gateway and observability fees, and non-token charges are excluded unless they appear in the committed record. Verify the provider source and quote the last-checked date with any copied figure, because model prices change often.
Hidden costs to watch
Flags attached to this source-backed record. They do not change the list price but they change the effective bill.
- No hidden-cost flags are attached to this source-backed record, but always budget for output-token blow-up, retries, and cache misses.
Frequently asked questions
How much does OpenAI gpt-5.4 cost per 1M tokens?
gpt-5.4 lists at $2.50 per 1M input tokens and $15.00 per 1M output tokens in the ByteCosts index, with cached input reads at $0.250 per 1M. Output is about 6.0x the input price, so completion length drives the bill. See the price card and cost examples above for current per-workload figures.
Is gpt-5.4 cheaper than the alternatives?
On combined list price, 3 records in the index come in lower than gpt-5.4, listed in the cheaper-alternatives table. Whether one is actually cheaper for you depends on your token mix, cache hit rate, and quality bar, so compare on your real workload in the calculator.
How fresh is this gpt-5.4 price and where is it from?
The price comes from OpenAI's official source, normalized into the ByteCosts pricing index, graded A+ confidence, and last checked 2026-06-27. Prices are list prices and exclude negotiated discounts; verify the provider source before production billing.
Cite this page
OpenAI gpt-5.4 API pricing. ByteCosts. Updated 2026-06-27. https://bytecosts.com/pricing/openai/gpt-5-4/