GPT-5.5 Released: Why OpenAI Calls This “A New Class of Intelligence
OpenAI announced GPT-5.5 on April 23, 2026, calling it “our smartest and most intuitive to use model yet.” It’s positioned as a flagship agentic model for coding, knowledge work, computer use, and early scientific research. A more powerful variant, GPT-5.5 Pro, is also rolling out. API pricing, as confirmed by Sam Altman, is $5 per 1M input tokens and $30 per 1M output tokenswith a 1M token context window.
Key Facts at a Glance
| Item | Detail | Source |
|---|---|---|
| Release date | April 23, 2026 | OpenAI |
| Variants | GPT-5.5, GPT-5.5 Pro | OpenAI |
| ChatGPT access | Plus, Pro, Business, Enterprise | OpenAI |
| GPT-5.5 Pro access | Pro, Business, Enterprise | OpenAI |
| Codex access | Available at launch | OpenAI |
| API access | “Very soon” — not yet live | OpenAI |
| API pricing | $5 / $30 per 1M input/output tokens | Sam Altman (OpenAI Dev Community) |
| Context window | 1 million tokens | Sam Altman (OpenAI Dev Community) |
| Latency | Matches GPT-5.4 per-token latency | OpenAI |
What OpenAI Says GPT-5.5 Does Better
According to OpenAI’s official announcement, GPT-5.5’s strongest gains are in four areas:
- Agentic coding — planning, tool use, and multi-step code work
- Computer use — navigating real software environments
- Knowledge work — spreadsheets, documents, presentations, research
- Early scientific research — biology, mathematics, data analysis
OpenAI emphasizes that GPT-5.5 uses “significantly fewer tokens” to complete Codex tasks than GPT-5.4, while maintaining the same per-token latency.
Verified Benchmark Results (from OpenAI’s Official Page)
These numbers are taken directly from OpenAI’s announcement. They represent OpenAI’s own reporting and should be read as vendor-published figures until independent leaderboards catch up.
| Benchmark | GPT-5.5 | GPT-5.4 | GPT-5.5 Pro | Claude Opus 4.7 | Gemini 3.1 Pro |
|---|---|---|---|---|---|
| Terminal-Bench 2.0 | 82.7% | 75.1% | — | 69.4% | 68.5% |
| Expert-SWE (Internal) | 73.1% | 68.5% | — | — | — |
| GDPval (wins or ties) | 84.9% | 83.0% | 82.3% | 80.3% | 67.3% |
| OSWorld-Verified | 78.7% | 75.0% | — | 78.0% | — |
| Toolathlon | 55.6% | 54.6% | — | — | 48.8% |
| BrowseComp | 84.4% | 82.7% | 90.1% | 79.3% | 85.9% |
| FrontierMath Tier 1–3 | 51.7% | 47.6% | 52.4% | 43.8% | 36.9% |
| FrontierMath Tier 4 | 35.4% | 27.1% | 39.6% | 22.9% | 16.7% |
| CyberGym | 81.8% | 79.0% | — | 73.1% | — |
OpenAI also cites additional knowledge-work scores: SWE-Bench Pro: 58.6%, τ2-Bench Telecom: 98.0% (no prompt tuning), FinanceAgent: 60.0%, OfficeQA Pro: 54.1%, and 88.5% on an internal investment-banking modeling eval.
⚠️ Editorial note: These are OpenAI-reported benchmarks. Independent verification from sources like Artificial Analysis and Terminal-Bench leaderboards may take days or weeks to reflect GPT-5.5. Readers should weigh vendor scores accordingly.
What Developers Are Saying (Verified Quotes)
All quotes below appear in OpenAI’s official announcement:
“The first coding model I’ve used that has serious conceptual clarity.” — Dan Shipper, Founder & CEO, Every
“It genuinely feels like I’m working with a higher intelligence, and there’s almost a sense of respect.” — Pietro Schirano, CEO, MagicPath
“Losing access to GPT-5.5 feels like I’ve had a limb amputated.”— Engineer at NVIDIA (anonymous early-access tester)
“GPT-5.5 delivers the sustained performance required for execution-heavy work… It’s more than faster coding—it’s a new way of working.” — Justin Boitano, VP of Enterprise AI, NVIDIA
Real World Use Cases at OpenAI
OpenAI disclosed several internal deployments that illustrate GPT-5.5’s capabilities:
- More than 85% of OpenAI employees use Codex weekly across software engineering, finance, communications, marketing, data science, and product
- Finance team used Codex to review 24,771 K-1 tax forms (71,637 pages), accelerating the task by two weeks vs. the prior year
- Comms team built a scoring and risk framework to automate low-risk speaking-request triage in Slack
- Go-to-Market team member automated weekly business reports, saving 5–10 hours per week
Scientific Research Highlight: A New Ramsey Numbers Proof
According to OpenAI, an internal version of GPT-5.5 with a custom harness helped discover a new proof of a longstanding asymptotic fact about off-diagonal Ramsey numbers — a result in combinatorics that was later verified in Lean, the mathematical theorem-proving language.
OpenAI presents this as evidence that the model can contribute original mathematical arguments, not just code or explanations.
(Note: this is OpenAI’s reporting; independent mathematical community review is expected to follow.)
Availability & How to Access GPT-5.5
GPT-5.5 Pricing
As confirmed by Sam Altman in the OpenAI Developer Community:
- Input: $5 per 1 million tokens
- Output: $30 per 1 million tokens
- Context window: 1 million tokens
- Altman noted that tasks will require fewer tokens than GPT-5.4, partially offsetting the per-token rate
How GPT-5.5 Fits Into OpenAI’s “Superapp” Vision
In a press briefing reported by TechCrunch, OpenAI President Greg Brockman framed GPT-5.5 as another step toward a unified “superapp” that combines ChatGPT, Codex, and an AI browser. Chief Scientist Jakub Pachocki added: “We see pretty significant improvements in the short term, extremely significant improvements in the medium term.”
Chief Research Officer Mark Chen highlighted drug discovery and scientific research as key emerging use cases.
What GPT-5.5 Means for Marketing Agencies and Businesses
At MPG ONE Agency, we’re tracking several implications worth monitoring:
- 🤖 Agentic marketing workflows — multi-step campaigns that plan, execute, and verify themselves
- 📊 Spreadsheet and document automation — GPT-5.5’s knowledge-work gains directly affect day-to-day operations work
- 🔎 SEO and Generative Engine Optimization (GEO) — as AI search engines like Perplexity, ChatGPT Search, and Google’s AI Overviews grow, content strategies must adapt
- 💬 Customer support automation — τ2-Bench Telecom performance suggests high-quality service automation is increasingly feasible
- 💻 Codex-powered internal tooling — lower cost of building custom business apps
Editorial caution for agencies: Vendor-reported benchmarks should always be validated in your own workflows before major investment. We recommend pilot testing GPT-5.5 against your specific use cases before migrating production workloads.
When was GPT-5.5 released?
April 23, 2026, by OpenAI.
Is GPT-5.5 available via API?
Not yet. OpenAI states API access is coming “very soon,” pending additional safeguards.
How much does GPT-5.5 cost in the API?
$5 per 1M input tokens and $30 per 1M output tokens, with a 1 million token context window (confirmed by Sam Altman).
What is GPT-5.5 Pro?
A higher-capability variant available to ChatGPT Pro, Business, and Enterprise users. OpenAI reports it outperforms GPT-5.4 Pro across benchmarks like BrowseComp (90.1%) and FrontierMath Tier 4 (39.6%).
Are the benchmarks independently verified?
Not yet. The numbers in this article come from OpenAI’s own announcement. Independent leaderboards like Artificial Analysis and Terminal-Bench will update over the coming days and weeks.
Did GPT-5.5 really help discover a new math proof?
According to OpenAI, yes an internal version helped produce a new proof about off-diagonal Ramsey numbers, verified in Lean. Independent review by the mathematics community is expected.
Final Words
GPT-5.5 represents OpenAI’s most significant model release of 2026 so far, with especially strong claims around agentic coding and computer use. However, as with every new flagship release, we recommend that brands:
- Wait for independent benchmarks before making large-scale migration decisions
- Pilot test GPT-5.5 in real workflows against GPT-5.4 and Claude Opus 4.7
- Monitor the API launch for production-grade deployments
Written By :
Valentina Morelli
General Manager – MPG ONE
