Infercom Model Deprecations

Infercom regularly updates our products with new open-source models and consequently retires those that have fallen in market relevance. This document outlines our deprecation process, provides guidance on migration, and maintains a historical record of deprecated models with their recommended alternatives.

Infercom Inference Service model deprecation process

Note

For assistance during the deprecation process, or any general questions, please contact our support team at support@infercom.ai.

Notifications

• Production models: Production models will receive a deprecation notice of two to three weeks at a minimum, depending on usage patterns and the availability of newer versions.
• Preview models: Preview models may graduate to Production based on popularity and usage or be deprecated.

Communication channels

All notifications will be delivered through at least one of the communication channels listed below:

• Email: Users of models scheduled for deprecation will be notified via email. The email will provide details about the deprecation timeline and available alternative models, including recommended replacements.
• Documentation: Model deprecation changes will be recorded on this documentation page, including the deprecation date and recommended replacement models.
• Infercom Inference Service Playground: Models scheduled for deprecation will be clearly labeled in the Infercom Inference Service Playground to inform users of their upcoming retirement.

Model migration options

When a model is scheduled for deprecation, users should evaluate and select either the recommended replacement model or another option from our cloud offerings that best fits your application’s requirements.

Model migration steps

To ensure a smooth transition from a deprecated model to a supported alternative, please follow the steps below:

1. Deprecation notice

   Be attentive to email notifications and Infercom Inference Service platform notifications regarding model deprecations.

2. Evaluate alternatives

   Review our documentation to identify and select a suitable replacement model from our offerings.

3. Implement and test

   Integrate your chosen model into your application, conduct necessary testing to ensure compatibility, and deploy your model into production. When migrating to new models, please note that you may need to adjust your prompts to optimize performance.

4. Complete your migration

   To avoid service disruptions, finalize the transition by fully migrating your workload and application to your selected alternative model before the scheduled deprecation date.

Upcoming deprecations

Note

No models are currently scheduled for deprecation.

---

Infercom Inference Service model deprecation record

Model ID	Deprecation date	Replacement model
MiniMax-M2.7	September 30, 2026	gpt-oss-120b or gemma-4-31B-it (EU-hosted, 128k context). MiniMax-M3 via Global Model Catalog
MiniMax-M2.5	July 16, 2026	MiniMax-M2.7 (192k context) or gemma-4-31B-it (vision-capable)
gemma-3-12b-it	June 18, 2026	gemma-4-31B-it (EU-hosted, vision-capable)
Qwen3-32B	April 3, 2026	Meta-Llama-3.3-70B-Instruct or DeepSeek-V3.2 via Global Model Catalog
Qwen3-235B	April 3, 2026	Meta-Llama-3.3-70B-Instruct or DeepSeek-V3.2 via Global Model Catalog
DeepSeek-R1-0528	April 3, 2026	DeepSeek-V3.2 via Global Model Catalog
DeepSeek-R1-Distill-Llama-70B	April 3, 2026	Meta-Llama-3.3-70B-Instruct via Global Model Catalog
DeepSeek-V3-0324	April 3, 2026	DeepSeek-V3.2 via Global Model Catalog
DeepSeek-V3.1-Terminus	April 3, 2026	DeepSeek-V3.1 (EU-hosted) or DeepSeek-V3.2 (Global)
Llama-4-Maverick-17B-128E-Instruct	April 3, 2026	Meta-Llama-3.3-70B-Instruct via Global Model Catalog
Meta-Llama-3.1-8B-Instruct	April 3, 2026	Meta-Llama-3.3-70B-Instruct via Global Model Catalog
Meta-Llama-3.3-70B-Instruct	March 12, 2026	MiniMax-M2.5 (EU-hosted replacement). Llama-3.3-70B remains available via Global Model Catalog.
DeepSeek-V3-0324-cb	January 23, 2026	DeepSeek-V3.1