AI Models. OpenAI (GPT)
Generalist / code / analysis / agents

OpenAI (GPT)

GPT-5.5 Pro/Instant, 1M+ context, agents, multimodal

GPT-5.5 (Pro/Instant, Apr-May 2026) is the current frontier model: agentic capabilities, 1M+ context, native multimodality and NVIDIA co-design. The provider publishes per-1M token pricing (input/cached input/output); cache rewards good context management. In parallel, gpt-oss (Apache 2.0, MoE) enables self-hosting. OpenAI does not train on business data by default (API/enterprise).

Verified: 2026-05-22

Purchase decision (when to choose / when to avoid)

Choose if...

  • You want a 'default' model for most tasks: content + analytics + code + agents.
  • You need a mature ecosystem (integrations, tools, stable API, versioning).
  • You're considering open-weight (gpt-oss) as an alternative for self-hosting in selected processes.

Avoid if...

  • You want to avoid vendor lock-in and have full weight control (prefer open-weight).
  • You have very restrictive data/region requirements — compare terms (Azure OpenAI / enterprise).

Cost in practice (scenarios)

SMB / marketing 1-10 people

Usually: ChatGPT Team + optional API for automation.

  • copy, analytics, presentations
  • moderate volume
Company 10-200 people

Enterprise if you need SSO, data policies and control.

  • multiple departments
  • central IT governance
Automations (API)

Cost depends on tokens; cache, batch and context shortening pay off.

  • RAG, webhooks, integrations
These are estimates/scenarios (not an invoice). Actual cost depends on context length, number of users, limits and retention policies.

Deployment / data / enterprise

Deployment channels

  • ChatGPT (team/enterprise)
  • OpenAI API
  • Azure OpenAI Service
  • Self-host: gpt-oss (open-weight)

Data policy

Training on data
API/Enterprise: not by default (check terms in sources).
Retention
Depends on service (API vs ChatGPT vs Azure).
Data residency
Depends on region and provider (OpenAI / Azure).
In companies, API or ChatGPT Enterprise is usually chosen for data policies.

Enterprise readiness

Admin
ChatGPT Team/Enterprise + admin, policies, workspace.
SSO/SCIM
Depending on plan (Team/Enterprise).
Audit
Depending on plan (Team/Enterprise).
DPA
Available in business/enterprise variants (check terms).
Certifications
Depending on product and agreement (check sources).
Largest ecosystem — fastest time-to-value, but watch costs at volume.

Best use cases

  • quality-critical tasks (reasoning, 1M+ context, agents) — GPT-5.5 Pro/Instant
  • ecosystem integrations (BI, Notion, Slack, Microsoft, Azure OpenAI) and long-horizon agent ops (MCP)
  • self-host open-weight (gpt-oss 120B/20B) when control is needed and MLOps competence exists.

Strengths

  • Very mature API, deprecation policy and documentation; widest integration ecosystem.
  • GPT-5.5: native multimodality, agentic capabilities, 1M+ context; enterprise: customer 'owns inputs/outputs'.
  • gpt-oss: Apache 2.0, MoE, good for reasoning/tool-use; TensorRT-LLM and ecosystem support.

Weaknesses / risks

  • No weights for closed models → vendor lock-in; cost at high volume; risk of behavior changes (versioning).
  • Data compliance depends on mode (enterprise/API); model lifecycle (deprecations, e.g. DALL·E 2/3 snapshots 05/12/2026).

Current models (examples)

  • GPT-5.5 Pro / Instant (Apr-May 2026) — flagship variants, 1M+ context, agentic, native multimodal.
  • gpt-oss (120B/20B) — open-weight, Apache 2.0, on-prem/cloud/edge; 'harmony' format.

Alternatives (if this model doesn't fit)

CL
Claude
safety + long context + agentic capabilities
GE
Gemini
GCP/Vertex, multimodality, ZDR in config
DE
DeepSeek
cheap reasoning/code at high volume