Model details
GPT-5.4
GPT-5.4 marks a turning point in the GPT family by consolidating capabilities that once required a specialized coding model. The previous generation, GPT-5.3 Codex, carried the entire weight of OpenAI's computer-use ambitions—scoring 64% on the OSWorld-Verified benchmark with midtask steering and Skills automation. GPT-5.4 absorbed every one of those innovations into the mainline model and pushed further, achieving 75% on the same benchmark, surpassing the 72.4% human expert baseline for the first time. This is not a narrow coding specialist; it is a generalist that can operate desktop applications, click through interfaces, and navigate workflows autonomously. The architecture supports five reasoning effort levels—none, low, medium, high, and xhigh—giving developers control over the depth-versus-cost tradeoff for each request. With vision, function calling, structured output, and prompt caching built in, the model is designed for complex, multi-step problem solving where accuracy matters more than raw speed.
What makes GPT-5.4 particularly significant is the convergence of two separate development paths—reasoning depth and coding execution—into a single model that matches or exceeds industry professionals on 83% of GDPval knowledge-work tasks. The model handles entire codebases or large document collections in a single request through its extended that quick-info value window, which also reduces token consumption in agentic workflows by improving tool search mechanisms. Developers can integrate it through the OpenAI SDK with the same response format, tool calling syntax, and streaming API as previous generations, ensuring backward compatibility. A Pro variant at higher rates is available for high-stakes tasks requiring maximum accuracy. For teams previously relying on specialized coding models, GPT-5.4 offers a migration path that does not sacrifice the capabilities they depended on while adding the breadth of a general-purpose reasoning model.
Quick Info
Powered by- Provider
- Merge Gateway
- Model key
- openai/gpt-5.4
- Release date
- Mar 5, 2026
- Last updated
- Mar 5, 2026
- Knowledge cutoff
- 2025-08-31
- Input modalities
- Output modalities
- Capabilities
Cost
- Input token cost
- $2.50
- Output token cost
- $15.00
Limits
- Input tokens
- 922,000 tokens
- Output tokens
- 128,000 tokens
- Context window
- 1,050,000 tokens
OpenCode
Model variants
Transparent token rates
Compare gpt pricing
Rates are shown per one million tokens. Combined means one million input plus one million output tokens.
Latest news about GPT-5.4
Videos about GPT-5.4
More models around GPT-5.4
This exact model name is also listed by 27 other providers.