GPT-5.5 "Spud": release date, pricing forecast, and what we actually know right now
OpenAI finished pretraining on March 24. Polymarket gives it 78% odds of shipping by April 30. Nobody knows if it will be called GPT-5.5 or GPT-6. Here's everything confirmed, what the leaked signals say, and how pricing might land based on OpenAI's history.
Photo by Scott Rodgerson on Unsplash
What we know for certain
- -Pretraining completed March 24, 2026. The Information first reported it; Sam Altman confirmed to employees.
- -Altman called it "a very strong model that could really accelerate the economy" and said release is "a few weeks" away.
- -Greg Brockman: "two years of research," "big model feel," "not incremental."
- -As of April 9, Spud is in safety evaluation at OpenAI. No official launch date set.
- -Polymarket: 78% probability of release by April 30, 95%+ by June 30.
Why "Spud"?
It's a potato. OpenAI uses informal internal labels - GPT-5.3-Codex went by "Spark" - and now we get a root vegetable. The name has no technical meaning and won't appear in any product.
The product name is genuinely unresolved. Brockman's framing - "big model feel," "two years of research," "not incremental" - points toward GPT-6. But OpenAI's naming has been chaotic this year (GPT-5.1, 5.2, 5.3-Codex, 5.3-Codex-Spark, 5.4, 5.4 Mini, 5.4 Nano). The deciding factor is probably benchmark magnitude: if SWE-bench Pro lands in the high 70s near Claude Mythos territory (77.80%), they'll call it GPT-6. Low 60s means GPT-5.5.
For this post I'll call it GPT-5.5, with the caveat that GPT-6 is a real possibility.
Release timeline signals
| Date | Signal | Source |
|---|---|---|
| March 24 | Pretraining complete. Altman tells employees: 'few weeks.' | The Information |
| March 24 | Brockman on Big Technology podcast: 'big model feel, two years of research, not incremental.' | Big Technology Podcast |
| April 9 | Spud is in safety evaluation. No official launch date set. | Axios |
| April 9 | OpenAI readying staggered cyber rollout - unclear if Spud or separate product. | Axios, Security Boulevard |
| April 14 | GPT-5.2-Codex and GPT-5.1-Codex-Mini retirements announced - standard pre-launch housekeeping. | OpenAI API changelog |
| April 16 (unconfirmed) | Leaked date. Not confirmed by OpenAI. | adam.holter.com |
The retirement of GPT-5.2-Codex and GPT-5.1-Codex-Mini is the signal I'd weight most. OpenAI clears old models before pushing new ones live - they did the same thing before GPT-5.4 launched March 5.
On rollout order, GPT-5.4 is the recent precedent: consumer ChatGPT first, then free tier with usage caps 2-4 weeks later, then enterprise API 4-6 weeks after the consumer launch. If Spud ships mid-April, most developers won't have API access before May.
Pricing forecast
OpenAI hasn't said anything about pricing. But three years of data is readable. Major architectural leaps launch higher, then drop. Capability expansions within a generation can go either way.
| Model | Input / 1M | Output / 1M | vs prior |
|---|---|---|---|
| GPT-4 (March 2023) | $30.00 | $60.00 | Launch |
| GPT-4 Turbo (Nov 2023) | $10.00 | $30.00 | -67% input |
| GPT-4o (May 2024) | $2.50 | $10.00 | -75% input |
| GPT-5.3-Codex | $1.25 | $10.00 | Coding specialist |
| GPT-5.4 (Mar 2026) | $2.50 | $15.00 | +100% input vs 5.3 |
GPT-5.3 to GPT-5.4 is the most relevant precedent. Input doubled when GPT-5.4 absorbed GPT-5.3's coding capabilities and added computer use on top. Brockman is saying Spud is a bigger leap than that.
If it ships as GPT-6, OpenAI will probably price it higher at launch - $5-8/MTok input - then drop it in Q3 or Q4 once the enterprise sales cycle plays out. GPT-5.5 naming suggests $3-4/MTok input, signaling the upgrade without drifting too far from GPT-5.4's $2.50 floor.
There's real counter-pressure here. DeepSeek V3.2 charges $0.28/MTok input and performs competitively on reasoning benchmarks. Claude Opus 4.6 cut its price 67% this year. Google's Gemini 3.1 Pro is $2/MTok with a 1M context window. The market is compressing, and OpenAI can't ignore that indefinitely even on a flagship.
Two pricing scenarios
| Model / Scenario | Input / 1M | Output / 1M | Context |
|---|---|---|---|
GPT-5.4 (current baseline) | $2.50 | $15.00 | 272K / 1M |
GPT-5.5 scenario Incremental naming, market-aware pricing | $3-4 | $15-20 | 1M+ (flat) |
GPT-6 scenario Major release, launch premium | $5-8 | $25-40 | 2M (rumored) |
GPT-5.5 Mini (estimate) Following GPT-5.4 Mini ratio | $0.50-1.00 | $2-4 | TBD |
Estimates based on OpenAI pricing history. No official pricing announced as of April 11, 2026.
The market it walks into
Since GPT-5.4 launched in March, the frontier market has moved. Claude Opus 4.6 dropped 67% in price. Gemini 3.1 Pro leads GPQA Diamond at $2/MTok. DeepSeek V3.2 sits at $0.28/MTok.
| Model | Input / 1M | Output / 1M | Context |
|---|---|---|---|
| DeepSeek V3.2 | $0.28 | $0.42 | 128K |
| Gemini 3.1 Flash-Lite | $0.10 | $0.40 | 1M |
| Gemini 3.1 Pro | $2.00 | $12.00 | 1M |
| Grok 4.20 Beta | $2.00 | $6.00 | 1M+ |
| GPT-5.4 | $2.50 | $15.00 | 272K/1M |
| Claude Opus 4.6 | $5.00 | $25.00 | 200K |
| GPT-5.5 (est.) | $3-8 | $15-40 | 1-2M |
The comparison I keep coming back to: Gemini 3.1 Pro scores 94.3% on GPQA Diamond at $2/MTok. GPT-5.4 Pro scores 94.4% at $30/MTok. Nearly the same reasoning capability at 15x the price. If GPT-5.5 launches at $8/MTok with Gemini 3.1 Pro sitting at $2 and near-identical reasoning scores, OpenAI is betting the launch premium holds for enterprise contracts before buyers notice the math.
For most production teams, the mini tier is where actual budget decisions happen. GPT-5.4 Mini at $0.40/$1.60 is the workhorse for high-volume applications. It launched the same day as the flagship at roughly 16% of the flagship input price. A $4/MTok flagship would put a Mini around $0.60-$0.80/MTok - still competitive against Claude Haiku and Gemini Flash.
The staggered rollout question
Axios reported April 9 that OpenAI is readying a staggered rollout for a model with advanced cybersecurity capabilities - invite-only for vetted security partners first, similar to how Anthropic handled Claude Mythos Preview. The piece carefully notes it's unclear whether this is Spud or a separate product.
This matters for API timing. If Spud's security research capabilities match Mythos-level output, OpenAI may apply the same logic: consumer ChatGPT launch first, enterprise API access delayed 4-8 weeks pending additional safety review with select partners. GPT-5.3-Codex triggered a Fortune investigation into "unprecedented cybersecurity risks" - OpenAI is paying attention to that criticism.
For developers planning around API access specifically: May at the earliest, June as the safer planning assumption if the cybersecurity rollout logic applies to Spud.
What to do while you wait
If you're on GPT-5.4 now, there's probably nothing to do yet. The gap between consumer launch and enterprise API access means you're not migrating next week regardless of when Spud ships.
For teams evaluating which frontier model to build on right now: Gemini 3.1 Pro at $2/MTok with a 1M context window is hard to argue against on cost-performance. Artificial Analysis tracks live quality-vs-price rankings if you want independent data. Claude Opus 4.6 leads on coding tasks. GPT-5.4 has the best computer use scores and the Codex integration if you're building desktop agents. We track all API rates on the pricing comparison page.
When GPT-5.5 pricing drops, we'll update this post and add it to the table immediately. Use the cost calculator to model what different price points would cost you at your actual token volume.
Sources
- - Sam Altman internal message via The Information, March 24, 2026
- - Greg Brockman, Big Technology Podcast, March 2026
- - Axios: "OpenAI readies cyber model rollout," April 9, 2026
- - Polymarket: "GPT-5.5 released by" prediction market
- - Adam Holter, "OpenAI Spud leaked April 16 release date," April 2026
- - OpenAI API pricing page (GPT-5.4 family)
- - Security Boulevard: "OpenAI readies rollout of cyber model," April 9, 2026
- - Fortune: "GPT-5.3-Codex cybersecurity risks," February 2026