GPT-5.5 Spud fact-check: agentic coding is documented, Spud is not confirmed

Claim-by-claim verdict

What is not confirmed: GPT-5.5 Spud as a public release

The safest reading is narrow: GPT-5.5 Spud is unconfirmed in the public OpenAI documentation reviewed for this fact-check. OpenAI’s model documentation points users to GPT-5.4 as the latest model guide, and the all-models page describes GPT-5.4 as suited for agentic, coding, and professional workflows. ^[1][67][74]

That is different from saying Spud cannot exist internally. The evidence here simply does not establish a public model release, model card, API endpoint, pricing page, or official OpenAI model guide for GPT-5.5 Spud. The strongest Spud-specific sources in this set are third-party: TokenMix says no official GPT-5.5 release date, model card, or API pricing has been announced, and India Today describes Spud as rumoured to be GPT-5.5. ^[11][13]

What OpenAI does document: GPT-5.4, GPT-5-Codex, and Codex

For the public model layer, the official documentation reviewed here centers on GPT-5.4 and Codex. OpenAI’s all-models page lists GPT-5.4 for agentic, coding, and professional workflows, while the GPT-5.4 model page calls it a frontier model for complex professional work. ^[67][68]

For software engineering, GPT-5-Codex is the clearest named coding model in this source set. OpenAI describes GPT-5-Codex as a version of GPT-5 optimized for agentic coding tasks in Codex or similar environments, available through the Responses API. ^[79] OpenAI’s code-generation guide also says to use Codex for out-of-the-box coding agents and frames Codex as OpenAI’s coding agent for software development. ^[81]

The agent layer: Responses API and Agents SDK

OpenAI’s agent-building materials describe the platform as a set of composable primitives: models, tools, state/memory, and orchestration. The same material identifies the Responses API as the flagship API designed for building powerful agents. ^[58]

The Agents SDK is the code-first route when an application owns orchestration, tool execution, approvals, and state. ^[52] OpenAI also has separate guidance for orchestration and handoffs, which supports multi-step or multi-agent coordination as a documented platform pattern rather than a capability that needs the unconfirmed Spud label. ^[53]

The tool layer: calling functions, searching the web, and changing code

The official tool surface is broad. Function calling is documented for calling developer-defined tools, and web search is documented as a tool used through the Responses API. ^[35][27]

OpenAI also documents file search, computer use, MCP/connectors, shell, apply patch, code interpreter, and skills. ^{[20][34][37][89][85][86][90]} Together, those tools support workflows where an agent can retrieve private context, call application functions, search the web, connect to external systems, use controlled compute environments, and apply code changes through documented interfaces. ^{[20][35][27][37][89][85][86]}

Practical guidance for developers

If the practical question is whether developers can build agentic coding and tool-orchestrated applications today, the answer from the official docs is yes. If the question is whether those workflows require GPT-5.5 Spud, the answer from this source set is no.

1. Start from documented model guidance, especially GPT-5.4 and GPT-5-Codex/Codex, rather than unconfirmed model names. ^{[67][74][79][81]}
2. Use the Responses API and Agents SDK when your application needs state, tool execution, approvals, and orchestration. ^[58][52]
3. Add tools based on the workflow: function calling for your own APIs, web search for live web retrieval, file search for private content, MCP/connectors for external systems, and shell, apply patch, or code interpreter where code execution or code changes are appropriate. ^{[35][27][20][37][89][85][86]}
4. Use OpenAI’s orchestration and handoff guidance when work needs to move across specialized steps or agents. ^[53]

How to describe Spud accurately

The most accurate phrasing is: GPT-5.5 Spud is unconfirmed in the public OpenAI docs reviewed here. Avoid treating claimed release dates, benchmark results, context windows, API availability, or pricing as established facts from this source set. TokenMix explicitly says no official GPT-5.5 release date, model card, or API pricing has been announced, and India Today frames Spud as rumoured to be GPT-5.5. ^[11][13]

Until OpenAI publishes official documentation under that name, the evidence-backed developer path is the documented stack: GPT-5.4, GPT-5-Codex/Codex, the Responses API, Agents SDK, and the official tool guides. ^{[1][67][74][79][81][58][52]}