GPT-5.4
Flagship OpenAI family for long-context reasoning, agent loops, and multimodal tool use.
- Input
- $1.25 / 1M tokens
- Output
- $10.00 / 1M tokens
- Cached read
- $0.13 / 1M tokens
- Cached write
- $0.63 / 1M tokens
- Batch discount
- 50%
- Source
- OpenAI API pricing
- Verified
- Apr 1, 2026 (High)
Capabilities
- Modalities
- textimage→text
- Capabilities
- reasoningwebSearchimageInputbatchSupportcodeExecutionpromptCachingfunctionCallingstructuredOutputs
- Strengths
- Long-context agent tasks, Reliable tool orchestration, Strong coding and reasoning
- Tradeoffs
- Higher output-token cost than mini families, Latency rises on deep reasoning traces
Official Links
Benchmark Coverage
| Benchmark | Version | Score | Date | Source | Notes |
|---|---|---|---|---|---|
| SWE-bench Verified | 2026-Q1 | 72.1 pass@1 | Mar 1, 2026 | OpenAI launch materials | Vendor-published result.; Not directly comparable across benchmark families. |
| GPQA Diamond | 2026-Q1 | 67.4 accuracy | Mar 1, 2026 | OpenAI launch materials | Use alongside source methodology notes. |
Release History
| Release | Alias | Lifecycle | Release Date | Deprecation | Shutdown | Summary |
|---|---|---|---|---|---|---|
| GPT-5.4 Preview | gpt-5.4-preview | Preview | Nov 20, 2025 | Mar 1, 2026 | — | Initial preview release for early adopters. |
| GPT-5.4 | gpt-5.4 | Active | Mar 1, 2026 | — | — | Stable family promoted to general availability. |
Host Coverage
| Host | Type | Context | Pricing Note | Differences |
|---|---|---|---|---|
| OpenAI API | first-party | 1.0M | Reference pricing and newest snapshots land here first. | Prompt caching; Batch API; Code execution tools |
| Azure OpenAI | cloud | 1.0M | Regional enterprise availability can lag the first-party API. | Enterprise regional controls; Separate billing and quotas |
Migration Guidance
Recommended upgrade path from GPT-4o for teams that need longer context and steadier tool execution.
Replacement models: gpt-4o, gpt-4o-mini
Breaking changes: Reasoning-heavy responses can take longer than GPT-4o., Prompt-caching strategy should be revisited to capture the lower cached-token rate.
Change Events
| Date | Type | Title | Description | Source |
|---|---|---|---|---|
| Nov 20, 2025 | preview_started | Preview launched | GPT-5.4 entered preview with long-context reasoning and agent tooling. | OpenAI changelog |
| Mar 1, 2026 | ga_released | General availability | The preview alias was promoted to the stable GPT-5.4 family with lower token pricing. | OpenAI changelog |
Other models from OpenAI
GPT-4.1, GPT-4.1 mini, GPT-4.1 nano, GPT-4o, GPT-4o mini, GPT-5.4 mini, GPT-5.4 nano