What Is GPT-5.5?
GPT-5.5 is OpenAI’s frontier model, released April 23, 2026 as the successor to GPT-5.4. It’s a fully retrained base model — the first since GPT-4.5 — designed for agentic workflows: autonomous, multi-step tasks across coding, computer use, and research rather than single-turn chat. With Claude Fable 5 currently suspended under a US export-control directive, GPT-5.5 has become one of the most-recommended fallback models for developers who need a working frontier-tier alternative today.
Key Features
- Agentic-first design — built to understand complex goals, use tools, verify its own work, and complete multi-part tasks with minimal supervision.
- 1 million token context window — via the API (922K input, 128K output); 400K in Codex.
- Native multimodal input — text and image inputs in a single system.
- Major long-context jump — MRCR v2 at 1M tokens more than doubled from GPT-5.4’s 36.6% to 74.0%.
- Token efficiency gains — roughly 40% fewer output tokens per Codex task than GPT-5.4, partially offsetting the price increase.
- Powers ChatGPT, Codex, and GitHub Copilot — available across ChatGPT Plus, Pro, Business, and Enterprise, plus OpenAI’s Codex coding environment.
Benchmarks
GPT-5.5 leads every publicly available model on Terminal-Bench 2.0 at 82.7% (using its native Codex CLI harness). On a level playing field using the standard Terminus-2 harness, the picture shifts: Claude Opus 4.8 leads on SWE-bench Pro (69.2% vs GPT-5.5’s 58.6%) and on the Artificial Analysis Intelligence Index (61.4 vs 60.2). GPT-5.5 holds its lead specifically on Terminal-Bench when measured with its own tooling — a genuinely strong result, just not a directly comparable one.
Pricing
GPT-5.5 costs $5 per million input tokens and $30 per million output tokens via the API — double GPT-5.4’s $2.50/$15 rate. OpenAI frames the effective increase as closer to 20% once token efficiency gains are factored in, though that holds mainly for agentic and coding workloads; on simple Q&A or single-turn content generation, the doubled price hits without the offsetting efficiency.
How It Compares
Against Claude Opus 4.8, GPT-5.5 is the better pick specifically for terminal-based agentic coding measured on its own harness, while Opus 4.8 wins more broadly on real-world coding benchmarks and costs less ($5/$25 vs $5/$30). Against Claude Fable 5 — currently unavailable everywhere following the June 12 export-control suspension — GPT-5.5 is one of the most commonly recommended substitutes precisely because it isn’t subject to the same restriction. See our Fable 5 alternatives breakdown for the full comparison.
Is GPT-5.5 Worth It?
Upgrade or adopt GPT-5.5 if your workload is agentic: autonomous coding, computer-use automation, or long-context document work, where the efficiency gains genuinely offset the price increase. Hold off if you’re running simple chatbots or single-turn content generation, where the doubled per-token cost lands without a matching efficiency benefit — GPT-5.4 or a cheaper model may serve you just as well at lower cost.
Last updated: June 20, 2026.