manifest 0.1.3: capabilities.thinking.control = never (what the bundle's template does about thinking, derived from the template inside the bundle)

Card: definition line (LiteRT, 2026-09)

manifest: same-device CPU control rows on the Galaxy S26 for the flagship GPU recommendation (S7 backfill 2026-09-05); the recommendation is rewritten from the measured pair (measured win, kept for prefill, or reversed to cpu)

Add litertlm_manifest.json (deployment manifest 0.1.2: variant selection, verified backends, Android/Apple recommendations, measured rows with conditions)

Card: base_model_relation: quantized (list under the base model's Quantizations)

Add measured Raspberry Pi 5 (CPU) section

Card: note the tokenizer-section fix (weights unchanged)

Replace the embedded tokenizer.json with the upstream one (weights, graph and metadata unchanged). The previous copy carried the GPT-2 default pre-tokenizer instead of the model's own regex, so digits and punctuation before newlines split differently from the upstream tokenizer on every turn; prompts now tokenize identically (see the card note).

Add measured Galaxy S26 GPU backend section

Card: measured Performance table (M4 Max CPU/GPU + on-device where measured), accuracy note

Card: add measured Performance table (M4 Max CPU/GPU + iPhone 17 Pro), accuracy note, verified one-command usage

Document Gallery 1.0.16 direct Hugging Face import + desktop LiteRT-LM CLI (serve/run)

docs: add Training data & PII section (submission-guideline compliance)

Upload folder using huggingface_hub

initial commit