GPT-5.5 Review 2026: Benchmarks, Pricing & What Changed
OpenAI released GPT-5.5 on April 23, 2026 — internally codenamed "Spud" and described as the first fully retrained base model since GPT-4.5. Here's what changed, what the benchmarks show, and whether it's worth the upgrade.
Reviewed 2026-07-05 · Source: OpenAI announcement
Quick Verdict
GPT-5.5 is the biggest jump in OpenAI's lineup since GPT-4.5 — a genuine base-model retrain rather than a fine-tune, with benchmark gains concentrated in professional knowledge work (finance modeling, customer-service workflows, computer use) rather than just chat quality. At $5/$30 per million input/output tokens it costs slightly more than GPT-5.4, but the 1M context window and GDPval/OSWorld gains make it the clear default choice for new agentic and knowledge-work projects in 2026.
What's New in GPT-5.5
First full base-model retrain since GPT-4.5
GPT-5.5 (internally codenamed "Spud") is a ground-up rebuild rather than an incremental fine-tune of GPT-5 or GPT-5.4. OpenAI describes it as its smartest and most intuitive model yet, and frames it as "the next step toward a new way of getting work done on a computer."
GPT-5.5 Instant
Alongside the Thinking and Pro variants, OpenAI shipped GPT-5.5 Instant — a faster, more personalized default model for ChatGPT aimed at everyday conversational use rather than deep reasoning chains.
Knowledge-work benchmarks over generic leaderboards
OpenAI leaned on occupation-specific evals (GDPval, FinanceAgent, OfficeQA Pro, investment-banking modeling) rather than only generic reasoning benchmarks, positioning GPT-5.5 as a model for professional deliverables rather than just chat quality.
1M token context window
GPT-5.5 ships with a 1 million token context window in the API, up from the shorter default context most GPT-5-series models launched with — useful for large codebases, long documents, and multi-file agentic tasks.
Benchmark Scores
OpenAI emphasized occupation-specific knowledge-work evals over generic leaderboards for this release.
| Benchmark | GPT-5.5 |
|---|---|
| GDPval (44-occupation knowledge work) | 84.9% |
| OSWorld-Verified (computer use) | 78.7% |
| Tau2-bench Telecom (customer service) | 98.0% |
| FinanceAgent | 60.0% |
| Internal investment-banking modeling | 88.5% |
GPT-5.5 Pricing
API pricing per million tokens, verified against OpenAI's current pricing page.
Input: $5 / million tokens
Output: $30 / million tokens
1M token context window. Available in the Responses and Chat Completions APIs.
Input: $30 / million tokens
Output: $180 / million tokens
Higher-accuracy variant for the hardest reasoning and agentic tasks.
Input: $2.50 / million tokens
Output: $15 / million tokens
Still available via API for cost-sensitive workloads; cached input at $0.25/M.
API model ID: gpt-5.5. Available via OpenAI Platform.
Who Should Upgrade to GPT-5.5?
Upgrade now if you're running:
- ✓ Finance, investment-banking, or spreadsheet-modeling agents
- ✓ Computer-use or browser automation workflows
- ✓ Customer-service agents needing high accuracy without prompt tuning
- ✓ Long-document or large-codebase tasks needing the 1M context window
Stay on GPT-5.4 if:
- → You're cost-sensitive — GPT-5.4 is nearly half the input/output price
- → Your workload is simple chat or summarization where model gains won't show up
- → You haven't re-run your evals yet — always benchmark on your own tasks before migrating production traffic
Frequently Asked Questions
What is GPT-5.5?
GPT-5.5 is OpenAI's flagship large language model, with GPT-5.5 Thinking and GPT-5.5 Pro released April 23, 2026, and API access following April 24, 2026. Internally codenamed "Spud," it's the first fully retrained base model since GPT-4.5 rather than an incremental update to GPT-5 or GPT-5.4. OpenAI also shipped GPT-5.5 Instant, a faster default model for everyday ChatGPT use.
How much does GPT-5.5 cost via the API?
GPT-5.5 is priced at $5 per million input tokens and $30 per million output tokens, with a 1 million token context window, available in both the Responses and Chat Completions APIs. GPT-5.5 Pro, the higher-accuracy variant, costs $30 per million input tokens and $180 per million output tokens. The previous-generation GPT-5.4 remains available at $2.50/$15 per million tokens for cost-sensitive workloads.
What are GPT-5.5's benchmark scores?
GPT-5.5 scores 84.9% on GDPval (knowledge work across 44 occupations), 78.7% on OSWorld-Verified (autonomous computer use), and 98.0% on Tau2-bench Telecom (customer-service workflows) without prompt tuning. It also scores 60.0% on FinanceAgent and 88.5% on internal investment-banking modeling tasks — benchmarks OpenAI chose specifically to demonstrate professional knowledge-work capability rather than only generic reasoning.
Is GPT-5.5 better than GPT-5.4?
Yes — GPT-5.5 is a full base-model retrain, not a fine-tune, and OpenAI positions it as a meaningful jump over GPT-5.4 in agentic and professional knowledge-work tasks specifically. GPT-5.4 remains available at a lower price point ($2.50/$15 per million tokens vs. GPT-5.5's $5/$30) for teams that don't need the latest model's gains.
How does GPT-5.5 compare to Claude Opus 4.8?
Both are current-generation flagship models released in 2026. GPT-5.5 leans into occupation-specific knowledge-work benchmarks (GDPval, FinanceAgent) and computer-use tasks, while Claude Opus 4.8 emphasizes coding honesty and agentic dynamic workflows in Claude Code. Pricing is close: GPT-5.5 is $5/$30 per million input/output tokens vs. Opus 4.8's $5/$25. The better fit depends on whether your workload is general knowledge work (GPT-5.5) or agentic coding (Opus 4.8).
What is GPT-5.5 Pro for?
GPT-5.5 Pro is a higher-accuracy variant priced at $30 per million input tokens and $180 per million output tokens, intended for the hardest reasoning and agentic tasks where the standard GPT-5.5 isn't accurate enough to justify the lower cost. It trades speed and price for maximum reliability.
Try GPT-5.5
Available today in ChatGPT and via the OpenAI API.
Try ChatGPT →API developers: model ID is gpt-5.5
Affiliate disclosure: Some links on this page are affiliate links. If you sign up through them, AISO Tools may earn a commission at no extra cost to you. This never affects our rankings or reviews.
📬 Get the best new AI tools delivered weekly
One concise email with fresh launches, trending picks, and featured standouts.
Join thousands of professionals who discover the best AI tools every week. No spam — unsubscribe anytime.