GPT-3.5 Turbo made role-based chat messaging an ordinary API building block. It powered many early assistants and business prototypes, but it is now a historical integration case: OpenAI’s model documentation records that the GPT-3.5 Turbo API was shut down on October 23, 2026, two days before this article’s cutoff.
Historical research cutoff: 25 October 2026. Primary-source historical claims are date-scoped; cloud entitlements, region-specific availability, current pricing and rate limits are not implied.
Release identity at a glance
- Provider
- OpenAI
- Model family
- GPT-3.5 Turbo
- Public launch
- 2023-03-01
- API ID / checkpoint
- gpt-3.5-turbo
- API retirement status
- 2026-10-23
The release and why it mattered
The March 2023 launch paired GPT-3.5 Turbo with a developer-facing chat endpoint; a later update clarified that “ChatGPT API” had been an introductory name and should be understood as the GPT-3.5 Turbo API. That terminology matters in documentation audits: a consumer ChatGPT release, the API service, and a particular dated model snapshot were never guaranteed to share the same lifecycle.
Capabilities and genuine use cases
Role-segmented messages offered clearer separation of system instructions, user input and assistant replies than a single completion prompt. Yet roles did not guarantee safe boundaries against hostile document content, nor did they enforce JSON validity. Basic classification, drafting and chat were legitimate workloads, provided outputs were validated before entering another system. In July 2024 OpenAI explicitly pointed developers toward GPT-4o mini as a successor.
Integration and version identity
To migrate, enumerate the exact model strings in source code, environment variables, saved assistants and request logs. Sample representative real prompts including long threads, non-English input, tool-routing instructions and strict parser cases. Compare error rates, latency and total token consumption rather than relying on a vendor headline benchmark. Keep the replacement choice separate from the decommissioned original in any publication archive.
Limits and evidence gaps
OpenAI’s current model page describes a 16,385-token context and 4,096-token maximum output for this legacy model; these values are documentation facts, not an entitlement that survives shutdown. The model’s 2026 retirement means a successful historic curl command cannot be treated as executable current documentation. Retention, geographic access and exact older snapshot dates need additional archival evidence.
How to choose or migrate
For readers diagnosing a failing 2023 app, the priority is removing hard-coded `gpt-3.5-turbo` calls, refreshing tests and preserving a timeline of the switch. A newer mini model may fit a lightweight workflow, but the decision belongs to current evaluation and AI 10’s dated price ledger, not a claim that old and new behavior are interchangeable.
Reproducible evaluation checklist
Build three distinct test sets: the regular path, an adversarially ambiguous path, and a regression set drawn from failures. Hold prompts and tool access constant while comparing snapshots. Never use a consumer product announcement to fill an API region or account-entitlement field.
If the historical route is retired, keep the example as an archived request only. For active migration, identify a current model from the provider catalogue and perform a staging replay before changing production configuration.
Sources & Further Reading
- openai.com — documentation/reference (supplier-provided reference; not independently checked in this pass)
- developers.openai.com — documentation/reference (supplier-provided reference; not independently checked in this pass)
- developers.openai.com — documentation/reference (supplier-provided reference; not independently checked in this pass)
Source documentation establishes vendor-published facts, not independently measured model quality, tenant access, contractual rights or current prices. Use a dated pricing ledger and an account-specific entitlement check for deployment decisions.