Current models and request compatibility
The gateway now has current request profiles for GPT 6.1 Sol, Claude Sonnet 5.5, Gemini 3.8 Flash, Mistral Medium 3.5 and Small 2603, and Kimi K2.7 and K3 variants. This update improves reasoning controls, sampling adaptation, and multi-turn tool translation. Access and quota still depend on the provider account behind your key.
GPT 6.1 Sol and GPT 6 Astra tool requests require the Responses API. Gemini 3.8 Flash supports low, medium, and high thinking; minimal maps to low. Claude Fable 5.1, Opus 5.5, Sonnet 5.5, and Opus 4.7 fix sampling, and thinking counts toward their output limit. Inspect x-tokensense-adapted when troubleshooting an adjustment.
Image and audio updates
n8n 0.1.20 adds GPT Image 2.5 Flare and Sunburst, current Gemini image models, FLUX 3, Mini TTS, and current GPT transcription choices. Google image requests use generateContent; the Images compatibility route currently accepts one square 1K Gemini image. FLUX 3 also requires one image. GPT transcription models require JSON output through TokenSense.
Image Generation and Speech-to-Text now expose pricingAvailable and billingReason alongside cost. Text-to-Speech exposes pricing metadata when the gateway supplies billing headers. Mini TTS binary audio can succeed without usage sufficient to calculate the bill, returning binary_audio_usage_unavailable. Missing image usage dimensions can also leave cost unavailable. Unknown cost is not zero cost.
npm published; Cloud approved, awaiting publication
Versions 0.1.19 and 0.1.20 were intentional releases. npm 0.1.20 is published for self-hosted installations. As checked on 4 October 2026, n8n has approved 0.1.20 for Cloud and says it will be available in the next release. Version 0.1.18 remains live on Cloud until n8n publishes the update. Approval is not publication, and we are not promising a publication date.
Existing node defaults and saved model selections are preserved. Self-hosted users can update the community package to 0.1.20 and test a copy of their workflow before enabling new choices. Cloud users can keep using their installed version while awaiting the update; the newer dropdown and modality behavior described here requires 0.1.20.
Retired models and historical costs
Confirmed retired models return model_retired rather than being silently redirected. Pricing history and stored model IDs remain intact. Choose and test a compatible replacement, including its request format. This rollout does not retroactively correct historical customer spend.
n8n 0.1.20 node reference · Installation and updates · Model compatibility troubleshooting