Overview
GPT-5.5 is OpenAI's flagship as of April 2026, launching on April 23 and arriving roughly four months after GPT-5.4. OpenAI calls it the smartest and most intuitive model yet, pitched as one that can take a messy, multi-part task and carry it through — planning, using tools, checking its work and navigating ambiguity until the job is done. The biggest gains land in agentic coding, computer use, knowledge work and scientific research.
It keeps a 1-million-token context window (400K in Codex) and matches GPT-5.4's per-token latency while performing at a higher intelligence level, using significantly fewer tokens for the same Codex tasks. Two tiers ship — GPT-5.5 and GPT-5.5 Pro — with a Codex Fast mode that generates tokens 1.5× faster, and it releases with OpenAI's strongest safeguards to date, including a focus on cyber defense.
Key capabilities
- 1-million-token context window (400K in Codex)
- GPT-5.5 and GPT-5.5 Pro tiers
- Codex Fast mode runs 1.5× faster
- State-of-the-art agentic coding and computer use
At a glance
Terminal-Bench 2.0
82.7%
SWE-Bench Pro
58.6%
GPQA Diamond
93.6%
Context
1M tokens
Pricing: API from $5 / 1M input and $30 / 1M output; Pro at $30 / $180.
Pros & cons
What we like
- Top-tier reasoning and long-context recall
- Best-in-class reliability and tool use
- Strong artifact and document output
Trade-offs
- Frontier pricing is hard to justify for everyday tasks
- Pro tier latency is high
- Gains over 5.2 are incremental rather than dramatic
The verdict
GPT-5.5 is the steady, sensible frontier — it doesn't reinvent the 5-series but makes it more reliable, more useful at long context and more dependable in agentic loops. It's the safe default for high-stakes work if you can absorb frontier pricing.
Best for: Frontier reasoning, long-context analysis and agentic tool use.
Frequently asked questions
How is GPT-5.5 different from GPT-5.4?
It matches GPT-5.4's per-token latency while performing at a higher intelligence level and using fewer tokens for the same Codex tasks. It ships with a 1M context window in GPT-5.5 and GPT-5.5 Pro tiers, and posts state-of-the-art agentic coding and computer-use scores — 82.7% on Terminal-Bench 2.0 and 58.6% on SWE-Bench Pro.
How much does GPT-5.5 cost?
The API is priced at $5 per million input tokens and $30 per million output tokens for GPT-5.5, with Batch and Flex at half the standard rate. GPT-5.5 Pro runs $30 / $180 per 1M tokens for the highest-accuracy tier.