OpenAI releases GPT-5.5 with faster token efficiency and improved context understanding

OpenAI released GPT-5.5 on April 23, 2026, with the company claiming the model handles tasks faster than GPT-5.4 while using significantly fewer tokens for operations at the same difficulty level.

According to OpenAI's published data, the efficiency gains stem from improved contextual understanding. The model shows improvements across coding, computer use, scientific research, and document creation tasks.

Performance claims

OpenAI's internal benchmarks indicate GPT-5.5 completes same-difficulty tasks faster than GPT-5.4 with lower token consumption. The company states this pattern holds across multiple task types.

For agentic coding specifically, OpenAI claims GPT-5.5 demonstrates better performance in:

• Maintaining context across large codebases
• Reasoning through ambiguous system failures
• Verifying assumptions using available tools
• Propagating changes through related code

The company reports similar improvements in computer use tasks, attributing the gains to better intent understanding. OpenAI also claims reduced latency for document creation and other typically time-intensive operations.

Availability and access

GPT-5.5 is available starting April 23, 2026 to:

• ChatGPT Plus, Pro, Business, and Enterprise users
• Codex users across the same tiers

GPT-5.5 Pro (presumably a higher-capability variant) is available to Pro, Business, and Enterprise tiers only—excluding Plus users.

API access is not yet available. OpenAI stated APIs require "different safeguards" before release, but provided no timeline. The company says internal and external teams have tested the model for safety issues.

What this means

The claimed token efficiency improvements could reduce operational costs for users on metered plans, though OpenAI has not disclosed pricing changes. The focus on contextual understanding suggests the model may handle multi-step tasks and long conversations more reliably than GPT-5.4.

The delayed API release is notable—developers typically receive access alongside consumer products. The cited need for "different safeguards" may indicate concerns about automated or high-volume usage patterns that don't apply to the ChatGPT interface. Without API access or independent benchmarks, the performance claims remain unverified by third parties.