OpenAI completes pretraining of 'Spud' model, Altman promises 'very strong' release in weeks
OpenAI has completed pretraining on a new model codenamed 'Spud,' according to an internal memo from CEO Sam Altman reported by The Information. Altman claims the company expects a 'very strong model' within weeks that can 'really accelerate the economy.' To free compute resources, OpenAI will shut down its Sora video generation app.
OpenAI has completed pretraining on a new AI model codenamed "Spud," CEO Sam Altman told employees in an internal memo, according to reporting by The Information. Altman claims the company expects to deliver a "very strong model" in "a few weeks" capable of "accelerating the economy."
"Things are moving faster than many of us expected," Altman wrote in the memo.
Strategic Reorganization and Product Changes
Alongside the Spud announcement, OpenAI is restructuring its product organization. Fidji Simo's division is being renamed "AGI Deployment," signaling a shift toward deployment-focused work.
To allocate compute resources for Spud and other priorities, OpenAI will shut down Sora, its text-to-video generation application. The decision reflects the company's reallocation of infrastructure toward new model development.
Spud's Intended Role
According to the memo, Spud may serve as the foundation for OpenAI's planned desktop "superapp"—a unified application combining ChatGPT, the coding agent Codex, and the browser product Atlas.
Competitive Context
The timing of Spud's announcement comes as OpenAI seeks to close gaps with Anthropic, which has gained traction with agent-based AI systems for business customers, particularly through Claude Code. OpenAI's own coding offerings—Codex and Frontier—are described as still playing catch-up in the agent space.
What This Means
Spud represents OpenAI's next major model iteration following GPT-4 level systems. The internal positioning as a desktop superapp suggests the company is prioritizing integrated, deployed products over standalone chatbot interfaces. The shutdown of Sora—a product that generated significant attention at launch—indicates aggressive reallocation of capital and compute toward large language model development. The timeline of "a few weeks" to release means potential availability in early April 2026, pending no delays. No technical specifications, pricing, or capability details have been disclosed.
No specific details about Spud's architecture, context window, parameter count, or training data have been confirmed. All claims derive from Altman's internal memo as reported by The Information.
Related Articles
OpenAI releases GPT-5.5-Cyber with 85.6% CyberGym score, surpassing restricted Anthropic model
OpenAI released an updated GPT-5.5-Cyber model that scores 85.6% on CyberGym, surpassing Anthropic's Mythos 5 (83.8%) — the same model that triggered Trump administration export controls. The release proceeds without the political pushback that forced Anthropic to restrict foreign national access.
Mistral Releases Voxtral TTS: 4B Parameter Text-to-Speech Model at $0.016 per 1k Characters
Mistral AI has released Voxtral TTS, a 4B parameter text-to-speech model supporting 9 languages including English, French, German, Spanish, Dutch, Portuguese, Italian, Hindi, and Arabic. The model achieves 70ms latency for typical inputs and can clone voices from as little as 3 seconds of audio, priced at $0.016 per 1,000 characters.
Mistral Releases Mistral 3 Family: 675B-Parameter Large 3 MoE and Three Edge Models Under Apache 2.0
Mistral has released Mistral 3, including Mistral Large 3—a sparse mixture-of-experts model with 41B active and 675B total parameters—and three Ministral 3 edge models (3B, 8B, 14B). All models are released under Apache 2.0 license with multimodal capabilities and are available today on multiple platforms.
Google releases Gemini 3.1 Flash Image, claims Pro-level quality at $0.50 per 1M tokens
Google has released Gemini 3.1 Flash Image, internally codenamed "Nano Banana 2," an image generation and editing model with a 131K context window. The model is priced at $0.50 per 1M input tokens and $3 per 1M output tokens.
Comments
Loading...