GPT-5.5 Instant
OpenAI's GPT-5.5 Instant is the low-latency variant of GPT-5.5, powering ChatGPT Voice with full-duplex interaction, agentic coding, and tool coordination.
Model Overview
Capabilities, design details, and architectural traits
GPT-5.5 Instant - the voice-serving variant of OpenAI's agentic frontier model
GPT-5.5 Instant is the serving configuration of GPT-5.5 used behind the scenes in ChatGPT Voice, including both Standard Voice Mode and the full-duplex GPT-Live system. When a spoken question requires search, reasoning, or deeper work, GPT-Live delegates the task to GPT-5.5 Instant and brings the result back into the conversation while maintaining conversational flow.
GPT-5.5 itself is positioned as OpenAI's model for real work. It is designed to accept messy, multi-part tasks and independently plan, use tools, check its own work, navigate ambiguity, and persist until a task is finished. Its gains are in agentic coding, computer use, knowledge work, and early scientific research.
| Trait | Detail |
|---|---|
| Voice delegation target | GPT-Live delegates search, reasoning, and agentic tasks to GPT-5.5 Instant during live conversations |
| Full-duplex compatibility | Serves as the background intelligence model for GPT-Live's continuous, simultaneous listen-and-speak architecture |
| Latency efficiency | Matches GPT-5.4 per-token latency in real-world serving while delivering higher intelligence |
| Token efficiency | Uses significantly fewer tokens than GPT-5.4 to complete the same Codex tasks |
| Safety | Released with OpenAI's safeguards, including advanced cybersecurity and biology testing |
Delegated intelligence in voice conversations
The GPT-Live architecture decouples continuous interaction from deeper work. GPT-Live handles real-time conversation while GPT-5.5 Instant executes background tasks such as web search, multi-step reasoning, and tool use, returning results without breaking the flow of dialogue.
Benchmark Performance
Independent evaluations · Artificial Analysis
Accuracy & Capability Details
Compare Models Side-by-Side
Evaluate specifications, pricing, and independent benchmark indices
| Model Details | |||
|---|---|---|---|
| General Info | |||
| Provider | OpenAI | Anthropic | Anthropic |
| Release Date | June 25, 2026 | September 22, 2026 | September 28, 2026 |
| Knowledge Cutoff | - | - | Jun 2026 |
| Context & Limits | |||
| Context Window | 1.1M Best Context Window | 1M | 1M |
| Pricing (per 1M tokens) | |||
| Input Pricing | $5 | $4 | $2 Best Input Pricing |
| Output Pricing | $30 | $20 | $10 Best Output Pricing |
| Modalities | |||
| Inputs | textimage | textimagefile | textimagefile |
| Outputs | text | text | text |
| Benchmarks (0-100) | |||
| Intelligence Index | 26.0 | 57.6 Best Intelligence Index | 56.0 |
| Coding Index | 39.4 | - | - |
| Agentic Index | 16.6 | - | - |
Humanity's Last Exam
Extremely difficult logical reasoning and knowledge.
Long Context Reasoning
Logical reasoning over long context windows.
SciCode Benchmark
Scientific coding and mathematical modeling.
Independent evaluation data provided by Artificial Analysis. To view the latest benchmarks and full details, visit their official site.
Explore more from OpenAI
Other models by OpenAI
Top AI Models
Leading alternatives by intelligence score
