openai-node 7.46–7.50: Five Releases, 12-Hour Caching and Service Tiers Land in the SDK (2026)
Five openai-node releases shipped between October 31 and November 3, carrying 12-hour Responses caching, project service tiers, workload identity providers, and a Live Responses expansion into the TypeScript layer.
openai-node 7.46.0 through 7.50.0 carry late-October's official changelog batch into the TypeScript type layer — five versions across two days (October 31 and November 3) per the npm registry, checked 2026-11-04.
Release-to-feature map
| Version | Published (npm) | Additions | Official changelog match |
|---|---|---|---|
| 7.46.0 | 10-31 | agent context compaction, agent session review and tool settings, experimental agent time awareness, image-edit markup overlays, organization external key management, SIP DTMF commands and delivery events, typed Realtime references, hosted tool usage exposure, agent-session metadata filtering (#3070 et al) | several October agent / Realtime entries |
| 7.47.0 | 10-31 | typed Responses WebSocket warmup (#3084) | Realtime/Live line |
| 7.48.0 | 11-03 | Live Responses delegation tools and choices expansion (with exhaustive-switch caveat); sparse WebSocket output facts retained (#3101/#3088) | Live API expansion |
| 7.49.0 | 11-03 | workload identity providers and federation rules (#3103) | no dedicated changelog entry yet |
| 7.50.0 | 11-03 | 12-hour Responses caching and cache-write usage (#3106); project service tier settings (#3105) | 10-20 twelve-hour caching; 10-21 service tiers |
The cleanest match is 7.50.0: the October 20 changelog entry introduces optional 12-hour prompt caching for the Responses API, and October 21 lets Platform settings choose which service tiers projects may use, with organization settings able to restrict tiers for all projects — SDK changes #3106 and #3105 land exactly those.
The one upgrade caveat
7.48.0 is the only release that asks for action: the upstream note reads "exhaustive TypeScript switches may need to handle new variants," so code that dispatches exhaustively over Live Responses tool types or choices needs new branches for the added variants. Everything else in these five releases is additive on the Features side, with Bug Fixes limited to ecosystem-test dependency bumps and a docs clarification.
Status and sourcing
The official changelog (fetched as .md on 2026-11-04) ends October at 10-30's GPT-6 Rosalind with no November entries as of the fetch; 7.49.0's workload identity has no dedicated changelog entry, so the SDK types remain its verifiable surface. Version and dist-tag facts come from the npm registry's same-day response.
Next steps
- openai-python 3.9 / 3.10 and openai-node 7.11 / 7.12: prompt cache diagnostics, API key expiration, and GPT Image 2.5
- OpenAI SDK adds Browser/Computer toolsets (1/4): BrowserToolset turns driver methods into Responses function tools
- GPT-6.1 Terra lands in the API: the model card, official pricing, and the GPT-6 family price table
Key points
- 7.50.0 (released 2026-11-03, npm checked 2026-11-04): api gains 12-hour Responses caching and cache-write usage (#3106) and project service tier settings (#3105) — matching the official changelog's October 20 '12-hour prompt caching' and October 21 'project service tiers' entries
- 7.49.0 (11-03): api adds workload identity providers and federation rules (#3103)
- 7.48.0 (11-03): expands Live Responses delegation tools and choices, with an upstream note that exhaustive TypeScript switches may need to handle new variants; also retains sparse WebSocket output facts (#3101/#3088)
- 7.46.0 and 7.47.0 both shipped 10-31: the former lands agent context compaction, agent session review, time awareness, image-edit markup overlays, external key management, SIP DTMF, typed Realtime references and more (#3070 et al); the latter adds typed Responses WebSocket warmup (#3084)
- Per the official changelog (developers.openai.com, fetched as .md on 2026-11-04), the newest October entry is 10-30 GPT-6 Rosalind and November has no entries as of the fetch — the 7.49/7.50 features all trace back to October entries
Frequently asked questions
Official references
Related articles
openai-python 3.49.0 and openai-node 7.55.0: typed async function tools for the Agents API
The 2026-11-10 SDK pair: python v3.49.0 and node v7.55.0 add typed support for the existing async setting on beta Agents function-tool shapes; python also fixes Microphone WAV encoding. Verified against tag sources.
Read articleopenai-python 3.48.0 and openai-node 7.54.0: agent session event-stream obfuscation padding and stream_options
The 2026-11-09 SDK pair: python v3.48.0 adds obfuscation switches for agent session streams (stream_options, include_obfuscation, 33 event types); node v7.54.0 adds rosalind types and scheduling retention. Verified against tag sources.
Read articleopenai-python 3.42.0 and openai-node 7.46/7.47: agent compaction and review settings, external keys, SIP DTMF
openai-python v3.42.0 and openai-node v7.46/7.47 (2026-10-31): agent compaction, review, and spend-cap settings; time awareness; external keys; SIP DTMF; typed WebSocket warmup; image markup overlays.
Read articleSubscribe to GPTMap Weekly
One email every Monday: curated OpenAI updates, deep dives, and best practices. No ads, unsubscribe anytime.
Submitting opens Buttondown in a new tab to confirm your subscription.