Recommended models
Pick a model for the job. When a T-Cloud model fits, it is the recommendation.
| Use case | Recommended | Alternative | Where it runs | Context | Notes |
|---|---|---|---|---|---|
| General chat and assistants | | GPT-OSS 120B | T-Cloud Germany (EU) | 1M · 128K | GLM 5.3, GPT-OSS 120B and Mistral Small 4 run on T-Cloud. GLM 5.3 has a 1M context window. Mistral Small 4 also accepts images. |
| Reasoning and complex tasks | | GLM 5.3 Flash | T-Cloud Germany (EU) | 1M | GLM 5.3 and GLM 5.3 Flash run on T-Cloud with a 1M context window. reasoning_ accepts none, high, and max on GLM 5.3, and low, high, and max on GLM 5.3 Flash. |
| Coding and agents | | GLM 5.3 Flash | T-Cloud Germany (EU) | 1M | GLM 5.3 has the catalog's coding tag and a 1M context window. |
| Image and document understanding | | Gemma 4 | T-Cloud Germany (EU) | 256K · 102.4K | Both accept text and image on T-Cloud. Mistral Small 4 does too. |
| OCR | | PaddleOCR-VL-1.6 | T-Cloud Germany (EU) | 32K · 128K | Both run on T-Cloud, in the Essential, Professional and Agentic plans. Context windows are 32K and 128K. |
| Embeddings and RAG | | Jina Embeddings v2 Base DE | T-Cloud Germany (EU) | 8K | BGE-M3, Jina Embeddings v2 Base DE, and Qwen3 Embedding 8B run on T-Cloud. Context windows are 8K, 8K, and 32K. |
| Audio | | Whisper Large v3 Turbo | T-Cloud Germany (EU) | 448 | Speech to text on T-Cloud. Each model has a context window of 448 tokens. |
| Routing and decisions | | DiffusionGemma 26B A4B (Preview) | T-Cloud Germany (EU) | 256K | Both are preview models on T-Cloud. DiffusionGemma accepts text and image and has a 256K context window. |
| Image generation | | — | Azure Sweden Central | 4K | GPT Image 2 runs on Azure Sweden. No image-generation model runs on T-Cloud. |
-
General chat and assistants
Recommended
GLM 5.3
Alternative GPT-OSS 120B
Where it runs T-Cloud Germany (EU)
Context 1M · 128K
Notes GLM 5.3, GPT-OSS 120B and Mistral Small 4 run on T-Cloud. GLM 5.3 has a 1M context window. Mistral Small 4 also accepts images.
-
Reasoning and complex tasks
Recommended
GLM 5.3
Alternative GLM 5.3 Flash
Where it runs T-Cloud Germany (EU)
Context 1M
Notes GLM 5.3 and GLM 5.3 Flash run on T-Cloud with a 1M context window.
reasoning_accepts none, high, and max on GLM 5.3, and low, high, and max on GLM 5.3 Flash.effort -
Coding and agents
Recommended
GLM 5.3
Alternative GLM 5.3 Flash
Where it runs T-Cloud Germany (EU)
Context 1M
Notes GLM 5.3 has the catalog's coding tag and a 1M context window.
-
Image and document understanding
Recommended
Qwen 3.6 35B FP8 Alternative Gemma 4
Where it runs T-Cloud Germany (EU)
Context 256K · 102.4K
Notes Both accept text and image on T-Cloud. Mistral Small 4 does too.
-
OCR
Recommended
LightOnOCR-3-4B Alternative PaddleOCR-VL-1.6
Where it runs T-Cloud Germany (EU)
Context 32K · 128K
Notes Both run on T-Cloud, in the Essential, Professional and Agentic plans. Context windows are 32K and 128K.
-
Embeddings and RAG
Recommended
BGE-M3 Alternative Jina Embeddings v2 Base DE
Where it runs T-Cloud Germany (EU)
Context 8K
Notes BGE-M3, Jina Embeddings v2 Base DE, and Qwen3 Embedding 8B run on T-Cloud. Context windows are 8K, 8K, and 32K.
-
Audio
Recommended
Whisper Large v3 Alternative Whisper Large v3 Turbo
Where it runs T-Cloud Germany (EU)
Context 448
Notes Speech to text on T-Cloud. Each model has a context window of 448 tokens.
-
Routing and decisions
Recommended
d3 (Preview)
Alternative DiffusionGemma 26B A4B (Preview)
Where it runs T-Cloud Germany (EU)
Context 256K
Notes Both are preview models on T-Cloud. DiffusionGemma accepts text and image and has a 256K context window.
-
Image generation
Recommended
GPT Image 2 Alternative —
Where it runs Azure Sweden Central
Context 4K
Notes GPT Image 2 runs on Azure Sweden. No image-generation model runs on T-Cloud.
Migrating from GPT, Claude, or Gemini?
A model that is leaving the catalog is listed with its date and replacement on Deprecations.