OpenAI's next flagship, internally codenamed 'Spud,' completed pretraining on March 24 and Polymarket gives it a 78% chance of release before April 30, 2026.
Seventy-eight percent. That is Polymarket's estimated probability, as of April 12, 2026, that OpenAI releases its next flagship model before the end of April — completing a product cycle so compressed that less than seven weeks would separate the new release from its predecessor, GPT-5.4, which launched on March 5.
The model, known internally at OpenAI as "Spud," completed pretraining on March 24, according to reporting by The Information corroborated by engineering posts from former OpenAI researchers. Whether it ships as GPT-5.5 or GPT-6 depends on benchmark performance: CEO Sam Altman told employees the model is "a very strong system that could really accelerate the economy," suggesting a generational improvement over the current product. If internal evaluations confirm that, OpenAI is expected to apply the GPT-6 designation.
openai · gpt-6 · gpt-5.5
The figures that frame the stakes are substantial. OpenAI's annualized revenue crossed $25 billion in Q1 2026, according to numbers the company shared with prospective IPO investors reviewed by the Financial Times in March. ChatGPT has approximately 900 million weekly active users, per OpenAI's own disclosures — larger by active-user count than any commercial software product in history except Microsoft Windows and Google Search. Venture capital poured $242 billion into AI companies in the first three months of 2026, representing roughly 80% of all global venture funding that quarter, according to PitchBook data published April 7. A GPT-6 launch would arrive into the most capital-saturated AI market ever recorded.
Continue reading to see the full article
“OpenAI's annualized revenue crossed $25 billion in Q1 2026, according to numbers the company shared with prospective IPO investors reviewed by the Financial Times in March.”
The strategic logic behind a rapid release is straightforward. Google's Gemini 3 Ultra, released in early April 2026, outperformed GPT-5.4 on several reasoning and multimodal tasks, according to the LMSYS Chatbot Arena leaderboard. Anthropic's Claude 4 Opus, released in late March, holds the top position on legal and scientific document analysis. Each week GPT-5.4 sits below those benchmarks on the leaderboard is a week that enterprise procurement officers note.
Key Takeaways
→openai: "Spud" is OpenAI's internal codename for its next flagship model, which completed pretraining on March 24, 2026.
→gpt-6: "Spud" is OpenAI's internal codename for its next flagship model, which completed pretraining on March 24, 2026.
→gpt-5.5: "Spud" is OpenAI's internal codename for its next flagship model, which completed pretraining on March 24, 2026.
→artificial intelligence: "Spud" is OpenAI's internal codename for its next flagship model, which completed pretraining on March 24, 2026.
A significant step-change from OpenAI would reset the competitive narrative at a moment when rivals are converging. The AI frontier is no longer a single company's property. Microsoft, Google, Anthropic, Meta, and Mistral AI all deploy capable models at commercially relevant price points; the margin between them narrows with each new release. For OpenAI, which built its brand identity on always sitting at the frontier, the cost of falling behind — even temporarily — is measured in contract renewals and developer adoption.
The speed of OpenAI's release cadence is compressing economics throughout the industry. Mistral AI, which raised €830 million in debt financing in March 2026, built its business model around efficient mid-tier models that offer lower cost per token than the frontier leaders. Each time a frontier lab's flagship drops in price or performance ceiling, the commercial rationale for mid-tier alternatives weakens. Several open-source model developers interviewed by TechCrunch in early April described the pace as "existential pressure" for smaller players without sovereign or enterprise contracts to fall back on.
Advertisement
The risk OpenAI's promotional framing tends to bury is reliability. GPT-5.4 launched with a documented tendency to over-refuse sensitive queries and to produce confident but inaccurate citations in legal and medical contexts, according to testing by Alignment Forum researchers published March 18. Altman acknowledged the issue at a San Francisco developer meetup on March 27 and said it was being corrected. Whether Spud resolves these alignment problems — or introduces new ones under greater capability — will not be known until external researchers have access to the deployed model and the adversarial testing begins in earnest.
openai · gpt-6 · gpt-5.5
The next key date is April 30 — both the close of the Polymarket prediction window and the date of OpenAI's previously announced developer conference in San Francisco. If the company follows its recent pattern of unveiling products at live events, a GPT-6 announcement could accompany Altman's keynote. For enterprises currently in contract renewal cycles with existing AI vendors, the intervening two weeks are likely to be a period of deliberate delay — waiting to see what arrives before committing.
#openai#gpt-6#gpt-5.5#artificial intelligence#chatgpt#sam altman#ai model release#ai industry#machine learning#generative ai#tech investment#openai revenue
What is OpenAI's "Spud" model and when will it be released?
"Spud" is OpenAI's internal codename for its next flagship model, which completed pretraining on March 24, 2026. Polymarket assigns a 78% probability of release before April 30. Whether it ships as GPT-5.5 or GPT-6 depends on internal benchmark results; CEO Sam Altman described it as "a very strong system that could really accelerate the economy."
How does OpenAI's new model compare to Google Gemini 3 and Anthropic Claude 4?
As of April 13, 2026, Google's Gemini 3 Ultra outperforms GPT-5.4 on several reasoning and multimodal benchmarks, per the LMSYS Chatbot Arena leaderboard. Anthropic's Claude 4 Opus leads on legal and scientific document analysis tasks. A significant GPT-6 release could shift these rankings; external validation typically takes several weeks after a model's public launch.
Is OpenAI profitable?
OpenAI's annualized revenue reached $25 billion in Q1 2026, according to figures shared with prospective IPO investors reviewed by the Financial Times. The company has not disclosed full profit or loss figures publicly. ChatGPT has approximately 900 million weekly active users as of April 2026.
