🔥 UPDATED — tokenizer refreshed & quantization improved:
layer-0 dense MLP and shared experts now preserved in BF16
Chat Template from: https://huggingface.co/sanjxz/Laguna-S-2.1-Agentic-Chat-Template-Jinja 🔥

---

[!IMPORTANT]

Linear-RFA 4-bit quant of poolside/Laguna-S-2.1

This is tcclaviger/Laguna-S-2.1-RFA_L — an RFA 4-bit quantization of Poolside's Laguna-S-2.1. The _L (linear-only) build quantizes the MLP and expert linear layers to 4-bit (IQ4_NL grid, group 16, Hadamard-16 rotation, asymmetric block-float scales) while keeping all attention, the router gate, the lm_head, the layer-0 dense MLP, and the shared experts in bf16. 71.2 GB total, 4.504 bits/weight. All credit for the model to Poolside; this repo only changes the numerics. The full original model card is preserved verbatim below.

Runtime: requires tcclaviger/vllm:latest — an RDNA 4 (gfx12xx) vLLM image and the only build with the RFA kernels; no other vLLM build loads these weights. Not validated on any other hardware at this time.

DFlash speculative decoding works with this RFA checkpoint. Pair it with the Laguna-S-2.1-DFlash draft model exactly as documented in the vLLM section below — the RFA quant does not touch the attention path, so the DFlash draft head runs unchanged.

**DRY has been incorportaed if desired into the docker image as has repetition blocking guard, two layers of protention. See the help for DRY parameters.

[!NOTE]

Evaluation results (this RFA quant)

Eval	Result
Throughput	80.6 tok/s out @ conc 1 (DFlash accept ~2.1); 615.7 tok/s out @ conc 50, ISL 512
WikiText-2 PPL	8.109 ± 0.043 (n_ctx 2048, llama.cpp-compatible)
Loglikelihood acc	arc_challenge 0.503 / arc_easy 0.769 / winogrande 0.658 / hellaswag 0.814 (acc_norm)
tool-eval-bench (no-think)	89/100 (full 69 scenarios)
GSM8K / MMLU / IFEval	90% / 60% / 95% (prompt-level)
Long-context code recall	0.980 overall (py 1.00 / js 0.93 / rs 0.99 / cpp 1.00); tool pass 93.5%
Hard Mode agentic (thinking-ON)	73/100 — 10 pass / 2 partial / 3 fail

Test harnesses: vllm bench serve (random dataset, saturation sweep) · llama.cpp-compatible WikiText-2 perplexity · lm-evaluation-harness (loglikelihood, local-completions) · tool-eval-bench v2 (69 scenarios + GSM8K/MMLU/IFEval + 15 Hard Mode scenarios) · codeneedle (positional recall, 4 corpora + tool pass).

Hardware: tcclaviger/vllm:latest, 4× AMD AI PRO R9700 GPUs (TP4), Ryzen 9 9950X, 256 GB DDR5-6000. (Not yet fully tuned for throughput.)

---

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

Laguna S 2.1

Laguna S 2.1 is a 118B total parameter Mixture-of-Experts model with 8B activated parameters per token, designed for agentic coding and long-horizon work. It sits between Laguna XS 2.1 (33B-A3B) and Laguna M.1 (225B-A23B) in the Laguna series and shares the family recipe: a token-choice router with softplus gating over 256 routed experts plus one shared expert, grouped-query attention, and interleaved full/sliding-window attention.

Highlights

• Mixed SWA and global attention layout: 48 layers in a 1:3 global-to-SWA ratio (12 global attention layers, 36 sliding-window layers, window 512), with softplus attention gating and per-layer-type rotary scales
• 1M context: 1,048,576-token context window
• Native reasoning support: interleaved thinking between tool calls, with per-request control via enable_thinking
• Speculative decoding: a trained DFlash draft model is available for lower-latency serving
• Quantized variants: FP8, NVFP4, INT4 and GGUF
• OpenMDW-1.1 license: Use and modify the model and associated materials freely for commercial and non-commercial purposes (learn more about OpenMDW)

Model overview

• Number of parameters: 118B total, ~8B activated per token
• Layers: 48 (12 global attention, 36 sliding-window attention)
• Experts: 256 routed (top-10) plus 1 shared expert
• Attention: grouped-query, 8 KV heads, head dim 128; per-head softplus output gating
• Sliding window: 512 tokens
• Context window: 1,048,576 tokens
• Vocabulary: 100,352 tokens (Laguna family tokenizer)
• Modality: text-to-text
• Reasoning: interleaved thinking with preserved thinking

Benchmark results

Benchmarks as of 21 July 2026. Laguna S 2.1 in bold; a dash (-) marks a benchmark a model was not evaluated on. Scores marked * are as reported by third parties: Terminal-Bench 2.1 and DeepSWE via Artificial Analysis, SWE Atlas via Scale AI's official leaderboard, and Toolathlon Verified via its official leaderboard. Full evaluation trajectories: trajectories.poolside.ai.

Usage

Laguna S 2.1 uses the same laguna architecture as Laguna XS 2.1, so the same engine integrations apply (vLLM, SGLang, Transformers, TRT-LLM, llama.cpp). At 118B parameters the BF16 checkpoint needs multiple GPUs (roughly 236GB of weights); quantized variants reduce this substantially.

vLLM

vllm serve \
    --model poolside/Laguna-S-2.1 \
    --tensor-parallel-size 4 \
    --tool-call-parser poolside_v1 \
    --reasoning-parser poolside_v1 \
    --enable-auto-tool-choice \
    --served-model-name laguna \
    --default-chat-template-kwargs '{"enable_thinking": true}'

[!NOTE] Optional: speculative decoding with DFlash. Pair with the Laguna S 2.1 DFlash draft model by adding --speculative-config '{"model":"poolside/Laguna-S-2.1-DFlash","num_speculative_tokens":7,"method":"dflash"}'.

SGLang

python -m sglang.launch_server \
  --model-path poolside/Laguna-S-2.1 \
  --tp-size 4 \
  --reasoning-parser poolside_v1 \
  --tool-call-parser poolside_v1 \
  --trust-remote-code

TRT-LLM

trtllm-serve poolside/Laguna-S-2.1 --trust-remote-code \
    --tool_parser poolside_v1 --reasoning_parser laguna

Note the flag names differ from vLLM's (--tool_parser, and the reasoning parser is laguna, not poolside_v1).

llama.cpp

GGUF conversions are available at poolside/Laguna-S-2.1-GGUF. 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 7 -fa on --jinja --port 8000

Controlling reasoning

Laguna S 2.1 has native reasoning support and works best with preserved thinking: keep reasoning_content from prior assistant messages in the message history. The model will generally reason before calling tools and between tool calls, and may stop reasoning in follow-up steps if prior thinking blocks are dropped.

Thinking is controlled per request via the chat template:

extra_body={"chat_template_kwargs": {"enable_thinking": False}}

or at the server level with --default-chat-template-kwargs '{"enable_thinking": true}'. For agentic coding use cases we recommend enabling thinking and preserving reasoning in the message history.

License

This model is licensed under the OpenMDW-1.1 License.

Intended and Responsible Use

Laguna S 2.1 is designed for software engineering and agentic coding use cases, and you are responsible for confirming that it is appropriate for your intended application. Laguna S 2.1 is subject to the OpenMDW-1.1 License, and should be used consistently with Poolside's Acceptable Use Policy. We advise against circumventing Laguna S 2.1 safety guardrails without implementing substantially equivalent mitigations appropriate for your use case.

Please report security vulnerabilities or safety concerns to security@poolside.ai.