organisation card: GGUF links

README.md

@@ -31,7 +31,13 @@ or a training step from the receipt.
   8/10-bit integer tensor-core products, reaching the same held-out loss across five seeds, with
   both training runs bound into receipt chains and a bit-exact replay segment. Parity, not
   lift, is the claim.
-- b-posit8 GGUFs of open models for the exact profile, as they are published.
+- **b-posit8 GGUFs** for the exact profile, one repo per model with the upstream licence on the
+  card: [SmolLM2-135M](https://huggingface.co/Anomly/SmolLM2-135M-Instruct-bposit8) (the one
+  INVAR's fixtures and vectors reference), [SmolLM2-1.7B](https://huggingface.co/Anomly/SmolLM2-1.7B-Instruct-bposit8),
+  [Qwen2.5-0.5B](https://huggingface.co/Anomly/Qwen2.5-0.5B-Instruct-bposit8),
+  [Llama-3.2-1B](https://huggingface.co/Anomly/Llama-3.2-1B-Instruct-bposit8),
+  [Gemma-3-270M](https://huggingface.co/Anomly/Gemma-3-270M-it-bposit8),
+  [Mistral-7B-Instruct-v0.3](https://huggingface.co/Anomly/Mistral-7B-Instruct-v0.3-bposit8).
 
 ## What we do not claim