Use on OpenRouter · Use on Vercel AI Gateway · Release blog post

Laguna S 2.1 GGUF

GGUF conversions of Laguna S 2.1 for llama.cpp, plus the DFlash speculative-decoding draft model. See the base model card for architecture details, license, and usage guidance.

Files

Serving

Serve with Poolside's llama.cpp fork, branch laguna, which carries full Laguna support including DFlash speculative decoding. (Base Laguna support is also in upstream review: ggml-org/llama.cpp#25165.)

git clone --branch laguna https://github.com/poolsideai/llama.cpp
cd llama.cpp && cmake -B build && cmake --build build -j

./build/bin/llama-server -m laguna-s-2.1-Q4_K_M.gguf --jinja --port 8000

# with DFlash speculative decoding:
./build/bin/llama-server -m laguna-s-2.1-Q4_K_M.gguf \
  -md laguna-s-2.1-DFlash-BF16.gguf \
  --spec-type draft-dflash --spec-draft-n-max 15 \
  --spec-draft-override-tensor '.*=CUDA0' \
  -fa on --jinja --port 8000

Context length

These GGUFs ship configured for a 262,144-token (256K) context window. This is the configuration we recommend for best output quality.

The weights are native 1M checkpoints: training included a long-context extension stage up to 1,048,576 tokens. To use more than 256K of context with llama.cpp, override the rope configuration at load time:

--ctx-size 1048576 --rope-scaling yarn --rope-scale 128 --yarn-orig-ctx 8192

You may experience quality degradation with the 1M configuration. If you use it, we recommend sampling with --temp 0.7 --top-p 0.95.

This release also corrects the embedded yarn_attn_factor metadata (now 1.0; llama.cpp derives the YaRN attention scaling internally).