← All models See what we measured ourselves
Delentia SLM (v0.4)
Base: Llama 3.1 8B
What the repository contains
The repository holds the merged weights (safetensors), quantised GGUF files, a Modelfile for Ollama, training configurations and the output of the author's own local benchmark script.
huggingface.co/Delentia/delentia-slm-jitna-v0.4Limits and what is not verified
- The figures in the model card come from the author's own local script and were not independently reproduced. Some are not meaningful as written: for example an adapter swap time of 0.0000 ms, and answers that took between 4 and 13 minutes to stream.
- The 'peer-reviewed' tag on the card is not accurate: a Zenodo upload is a publication, not a peer review.
- Delentia's agent runtime does not use this model today. Its default is the model the owner sets, Ollama qwen2.5:7b unless changed. No evaluation of this model inside the runtime has been run.