Language Models


Name Total Parameters Quantization Context Length Available until More information
Kimi-K2.6 Total: 1.1T / Activated: 32B BF16 256,000 tokens 14/08/26 https://huggingface.co/moonshotai/Kimi-K2.6
Qwen3.6 36B BF16 262,144 tokens 14/08/26 https://huggingface.co/Qwen/Qwen3.6-35B-A3B
Llama-3-3 70B FP8 131,072 tokens 14/08/26 https://huggingface.co/meta-llama/Llama-3.3-70B-Instruct
Qwen3 32B BF16 30,000 tokens 14/08/26 https://huggingface.co/Qwen/Qwen3-32B
Gemma4 30.7B BF16 65,536 tokens 14/08/26 https://huggingface.co/google/gemma-4-31B-it
Latxa-Qwen3-VL 32B BF16 65,536 tokens 14/08/26 https://huggingface.co/HiTZ/Latxa-Qwen3-VL-32B-Instruct

Public models are available to the entire SCAYLE user community with active permissions to use this service. Each model included in this category is offered with an initial availability cycle of 90 days. Once this period has elapsed, the SCAYLE team will review its usage, performance, and demand to decide whether it should be maintained, renewed, or replaced with another alternative. Users are advised to consult this section periodically to check the updated status of each model.

Private models are restricted-access resources that are not shown by default in the general catalogue. To access them, they must be explicitly requested from the SCAYLE team via or through a ticket in GLPI. Permission granting is a manual process assessed by the administrators, who will decide whether access is appropriate based on resource availability and the requirements of the project. If, after receiving access confirmation, you do not see the model in your Caléndula environment or in the web interface, please notify .