Spud is the internal codename for OpenAI's next frontier model, expected to be branded GPT-5.5 or GPT-6. It completed pre-training in late March 2026 and is currently undergoing safety evaluations before public release. The model features trillions of parameters, native multimodal input, and a unified architecture that merges chat, coding, and agentic task execution into a single system.

How to access OpenAI Spud?

Overchat AI will offer Spud as soon as it becomes publicly available through OpenAI's API. Select OpenAI Spud from the model picker and start chatting — no separate OpenAI subscription needed. Basic access requires no account; sign up free for conversation history and longer sessions.

Who made OpenAI Spud?

OpenAI, the San Francisco-based AI lab founded in 2015 by Sam Altman, Greg Brockman, Ilya Sutskever, and others. OpenAI created the GPT series and ChatGPT, and is currently valued at over $100 billion following a record $122 billion funding round in early 2026. Spud represents the company's biggest bet since GPT-4.

OpenAI Spud (GPT-6) — Chat with OpenAI's Next Frontier AI Model

Q: How to use OpenAI Spud?

Give it a goal, not a micro-instruction. Spud is built to handle ambiguous, multi-step requests — describe the outcome you want and let the model figure out the path. Use follow-up messages to steer, narrow scope, or ask it to try a different approach.

What is OpenAI Spud?

Spud is OpenAI's codename for their next frontier language model — a multimodal system with trillions of parameters that completed pre-training in March 2026. Sam Altman told employees it could "really accelerate the economy." On Overchat AI, you'll be among the first to chat with Spud as soon as it launches publicly.

How to Use OpenAI Spud

Give Spud a real task

Give Spud a real task
Open the chat widget above and type what you actually need done — write a marketing brief, debug a Python script, summarize a 50-page PDF. Spud is built for complex, multi-part requests.

Spud delivers a structured result

The model chains reasoning steps internally before answering. You get a thought-through response — not a first draft, but something you can act on right away.

Keep pushing

Keep pushing
Spud holds context across long sessions, so follow-ups actually build on what came before. Refine a code solution, drill into a research topic, or shift direction completely. Sign up free to save your history and access additional models.

Get Started

What is OpenAI Spud?

Spud is the internal codename for OpenAI's next flagship language model. It finished pre-training around March 24, 2026, and is widely expected to launch as either GPT-5.5 or GPT-6 — though OpenAI has not confirmed the final branding. What is confirmed: CEO Sam Altman told staff the model could "really accelerate the economy," and the company shut down Sora entirely to redirect compute toward getting Spud ready for release.

The architecture behind Spud is multimodal from the ground up — trillions of parameters trained on quadrillions of tokens spanning text, images, audio, and potentially video. OpenAI employees have hinted that the model contains a capability that is "very different from what we've seen before," though no specifics have been shared. The design goal is a unified system where a single session can move between conversation, code generation, web research, and autonomous task execution without switching tools.

The timing is notable. Just days before Spud completed training, mathematician Terence Tao published a paper crediting ChatGPT Pro with generating part of a formal proof — the clearest signal yet that frontier models are becoming genuine research tools, not just writing assistants. Spud is expected to push that boundary further. Safety evaluations and red-teaming are underway, and a public release could come as early as April 2026.

OpenAI Spud Features

Spud brings stronger chain-of-thought reasoning, building on what OpenAI pioneered with the o-series. The model thinks longer before it speaks, producing answers that hold up under scrutiny rather than sounding plausible but falling apart on closer inspection.

Multimodal understanding is native, not bolted on. Spud processes images, audio, and text within the same context window, so you can drop in a screenshot of an error log, a photo of a whiteboard, or an audio clip and get a coherent response that accounts for all of it.

On the agentic side, Spud is designed to power OpenAI's unified platform — combining what used to be separate products (ChatGPT for conversation, Codex for coding, Atlas for web browsing) into one seamless session. The model can plan multi-step workflows, call tools, browse the web, and write code without handing off to a different system.

OpenAI Spud will be available on Overchat AI alongside Claude Mythos and DeepSeek V4. Try all four in one interface and decide which model handles your specific workload best.

OpenAI Spud Benchmarks

OpenAI has not released official benchmark scores for Spud yet — the model is still in safety evaluation and red-teaming. However, based on the trajectory from GPT-4o to GPT-5.2 and the scale of compute invested, industry analysts expect significant jumps in coding (SWE-bench), mathematical reasoning (MATH, GSM8K), and agentic task completion. We will update this section as benchmark data becomes available.