Synapse1.5-1B-GGUF

Synapse 1.5 1B

Synapse 1.5 is SamanyoluAI's compact edge model — a 1.2B-parameter instruction-tuned LLM packaged in Q8_0 GGUF for local deployment via LM Studio, llama.cpp, and other GGUF runtimes.

Synapse 1.5 is built on LiquidAI/LFM2.5-1.2B-Instruct, released under the LFM Open License v1.0.

Specs

Usage

LM Studio — search for Synapse1.5-1B-GGUF, or place the .gguf under ~/.lmstudio/models/SamanyoluAI/Synapse1.5-1B-GGUF/.

llama.cpp

llama-cli -m Synapse1.5-1B-Q8_0.gguf -p "Hello"

Python (huggingface_hub)

from huggingface_hub import hf_hub_download

path = hf_hub_download(
    repo_id="SamanyoluAI/Synapse1.5-1B-GGUF",
    filename="Synapse1.5-1B-Q8_0.gguf",
)

Changes from the base model

This release modifies metadata only — the weights are unmodified. The tensor data section is byte-identical to the base model (SHA-256 3402faf5...).

1. GGUF general.* metadata rebranded: name, basename, size_label, plus added version, organization, description, and license.
2. Chat template extended with a default Synapse identity, injected only when the caller supplies no system prompt. A caller-provided system prompt overrides it.

Architecture, tokenizer, vocabulary, and all lfm2.* hyperparameters are unchanged.

License

Released under the LFM Open License v1.0, inherited from the base model. Note two conditions carried over from that license:

• §5 limits commercial use to entities under $10M USD annual revenue.
• §7 grants no trademark rights. This release is not affiliated with or endorsed by Liquid AI, Inc.

Acknowledgements

Base model by Liquid AI. Packaged and released by SamanyoluAI.