poolside / Laguna-S-2.1-INT4

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

Laguna S 2.1-INT4

Laguna S 2.1-INT4 is a 117.6B total parameter Mixture-of-Experts model with 8.5B activated parameters per token designed for agentic coding and long-horizon work on a local machine. It uses Sliding Window Attention with per-head gating in 36 out of 48 layers for fast inference and low KV cache requirements.

Highlights

• Mixed SWA and global attention layout : Laguna S 2.1 uses softplus gating with per-layer rotary scales, enabling mixed SWA (Sliding Window Attention) and global attention layers in a 3:1 ratio (across 48 total layers)
• KV cache in FP8 : KV cache quantized to FP8, reducing memory per token
• Native reasoning support : Interleaved thinking between tool calls with support for enabling and disabling thinking per-request
• Local-ready : At 117.6B total parameters and 8.5B activated, the INT4 weights are roughly 72 GB. Available on Ollama and llama.cpp (BF16 and Q4_K_M only)
• 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

• Training: pre-training, post-training and reinforcement learning stages
• Number of parameters: 117.6B total with 8.5B activated per token
• Optimizer: Muon
• Layers: 48 layers (12 layers with global attention, 36 layers with sliding window attention)
• Experts: 256 experts with 1 shared expert
• Sliding Window: 512 tokens
• Modality: text-to-text
• Context window: 262,144 tokens
• Reasoning support: 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

Context length

This checkpoint ships 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, and quantization was calibrated at the 1M configuration. If you need more than 256K of context, restore the 1M configuration by editing config.json :

"rope_parameters": {
  "full_attention": {
    "factor": 128.0,
    "attention_factor": 1.4852030263919618
  }
},
"max_position_embeddings": 1048576

You may experience quality degradation with the 1M configuration. If you use it, we recommend sampling with temperature <= 0.7 and top_p <= 0.95 .

Recommended sampling

For the best balance of quality and reliability we recommend sampling with temperature 0.7 and top_p 0.95 .

Local deployment

Laguna S 2.1-INT4 is supported in vLLM and Transformers. Use Laguna S 2.1 with Ollama (with MLX support) or Llama.cpp (BF16 and Q4_K_M only) for the best results on your local machine.

vLLM

Serving requires vLLM 0.25.0 or later. See the main Laguna S 2.1 model card for the full recipe.

The full vLLM recipe is on the main Laguna S 2.1 model card and on the vLLM recipes page . Quantization is detected automatically from quantization_config in this checkpoint, so the same command works with poolside/Laguna-S-2.1-INT4 substituted for the model ID. No extra flags required.

INT4 support landed in vLLM with vllm-project/vllm#47154 . This checkpoint mixes INT4 and group-quantized INT8 experts, which earlier builds rejected in the Marlin WNA16 MoE path. Use a vLLM build that includes this fix.

Optional: speculative decoding with DFlash. Pair with the quantization-matched draft model poolside/Laguna-S-2.1-DFlash-INT4 by adding --speculative-config '{"model":"poolside/Laguna-S-2.1-DFlash-INT4","num_speculative_tokens":15,"method":"dflash"}' to the serve command.

SGLang

The full SGLang recipe is on the main Laguna S 2.1 model card and on the SGLang cookbook . SGLang reads the checkpoint quantization_config, so you can use the same launch command after replacing the model ID

Transformers

The full Transformers recipe is on the main Laguna S 2.1 model card . Substitute poolside/Laguna-S-2.1-INT4 for the model ID; quantization is detected automatically from quantization_config .

llama.cpp

Serve the GGUF builds with Poolside's llama.cpp fork, branch laguna . See the main Laguna S 2.1 model card for the build recipe. llama.cpp serves the BF16, Q8_0, and Q4_K_M GGUF builds, not the INT4 safetensors.

Ollama

Available on the Ollama library .

Controlling reasoning

Laguna S 2.1-INT4 uses the same reasoning controls (interleaved thinking, preserved reasoning, and the enable_thinking flag) as the base model. See the Controlling reasoning section of the main Laguna S 2.1 model card.

License

This model is licensed under the OpenMDW-1.1 License .

Intended and Responsible Use

Laguna S 2.1-INT4 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-INT4 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-INT4 safety guardrails without implementing substantially equivalent mitigations appropriate for your use case.

Please report security vulnerabilities or safety concerns to [email protected] .