Dedicated capacity
Supported models for Dedicated Capacity
Every open-weight model Infercom can deploy on dedicated capacity in Munich or on-premises, from GLM-5.3 and MiniMax M3.1 Flash to gpt-oss and Gemma 4.
These models can run on dedicated capacity in our data center in Munich, and on hardware in your own data center. Availability depends on the platform release. Confirm a model with us before you plan on it.
You choose the models for your capacity. Several models can run together in one bundle, depending on the models. See Model bundles.
Featured models
Section titled “Featured models”GLM-5.3(Z.ai): a mixture-of-experts model with about 750B parameters, about 40B active. 1M token context, text.MiniMax-M3.1-Flash(MiniMax): a mixture-of-experts model with 427B parameters, 23B active. 1M token context, text and image input. Built with MiniMax M3.1-Flash.
All models
Section titled “All models”| Model ID | Maker | Type | Max context | Status | Custom checkpoints | View on Hugging Face |
|---|---|---|---|---|---|---|
GLM-5.3 |
Z.ai | Text | 1M | Preview | No | Model card |
MiniMax-M3.1-Flash |
MiniMax | Text + image | 1M | Preview | No | Model card |
MiniMax-M3 |
MiniMax | Text + image + video | 1M | Preview | No | Model card |
MiniMax-M2.7 |
MiniMax | Text | 192k | Stable | Yes | Model card |
MiniMax-M2.5 |
MiniMax | Text | 160k | Stable | Yes | Model card |
gpt-oss-120b |
OpenAI | Text | 128k | Stable | Yes | Model card |
gpt-oss-20b |
OpenAI | Text | 128k | Preview | Yes | Model card |
Whisper-Large-v3 |
OpenAI | Speech-to-text | - | Stable | No | Model card |
gemma-4-31B-it |
Text + image + video | 256k | Stable | No | Model card | |
gemma-3-27b-it |
Text + image | 128k | Preview | No | Model card | |
gemma-3-12b-it |
Text + image | 128k | Preview | No | Model card | |
DeepSeek-V3.2 |
DeepSeek | Text | 128k | Stable | Yes | Model card |
DeepSeek-V3.1-Terminus |
DeepSeek | Text | 128k | Stable | Yes | Model card |
DeepSeek-V3.1 |
DeepSeek | Text | 128k | Stable | Yes | Model card |
DeepSeek-R1-0528 |
DeepSeek | Text | 128k | Stable | Yes | Model card |
DeepSeek-V3-0324 |
DeepSeek | Text | 128k | Stable | Yes | Model card |
DeepSeek-R1-Distill-Llama-70B |
DeepSeek | Text | 128k | Stable | Yes | Model card |
Llama-4-Maverick-17B-128E-Instruct |
Meta | Text + image | 128k | Stable | No | Model card |
Meta-Llama-3.3-70B-Instruct |
Meta | Text | 128k | Stable | Yes | Model card |
Meta-Llama-3.1-405B-Instruct |
Meta | Text | 16k | Stable | Yes | Model card |
Meta-Llama-3.1-70B-Instruct |
Meta | Text | 32k | Stable | Yes | Model card |
Meta-Llama-3.1-8B-Instruct |
Meta | Text | 16k | Stable | Yes | Model card |
Mistral-Large-3-675B-Instruct-2512 |
Mistral AI | Text | 32k | Preview | Yes | Model card |
qwen3-tts |
Alibaba Cloud | Text-to-speech | - | Preview | No | Model card |
Qwen3-235B-A22B-Instruct-2507 |
Alibaba Cloud | Text | 128k | Stable | Yes | Model card |
Qwen3-32B |
Alibaba Cloud | Text | 32k | Stable | No | Model card |
Llama-3.3-Swallow-70B-Instruct-v0.4 |
Tokyotech-llm | Text | 128k | Stable | Yes | Model card |
E5-Mistral-7B-Instruct |
intfloat | Embeddings | 4k | Stable | No | Model card |
- A Preview model is an early release. It can lack features, and it is not guaranteed to become Stable.
- Custom checkpoints shows whether you can bring your own fine-tuned checkpoint of the model. See Custom checkpoints.
- Read the licence on the model card before you use a model commercially. The MiniMax models have commercial terms.