Today was the perfect day for Poolside to drop Laguna S 2.1 because I just got these in! Finally have a half decent amount of VRAM. 3x V620 = 96 GB.
Summary
Poolside released Laguna S 2.1, and the user acquired 3x AMD V620 GPUs totaling 96 GB VRAM.
Similar Articles
Laguna S 2.1 looping fix incoming
Poolside released Laguna S 2.1, their most capable model for long-horizon tasks, along with multiple quantized variants (FP8, NVFP4, INT4, DFlash, GGUF) on Hugging Face.
Laguna-S-2.1 runs on my 6 years old gaming PC!
Laguna-S-2.1, a quantized AI model, runs on a 6-year-old gaming PC with an RTX 3080, achieving 10 t/s decode and using 8.3 GB VRAM and 52.2 GB host RAM.
Laguna S 2.1 Released: Cheaper than Deepseek v4 Flash, Better than V4 Pro
Poolside releases Laguna S 2.1, a 118B MoE model with 8B activated parameters per token, optimized for agentic coding. It claims to outperform DeepSeek V4 Pro while being cheaper than DeepSeek V4 Flash, with a 1M context window and open-source license.
@pupposandro: PFlash now run @poolsideai's Laguna-XS.2 (33B-A3B MoE) on a single RTX 3090. - 111 tok/s decode @ short ctx - 128K TTFT…
PFlash now supports running @poolsideai's Laguna-XS.2 (33B-A3B MoE) on a single RTX 3090, achieving 111 tok/s decode and 5.4x faster prefill than llama.cpp, with NIAH passes up to 131K context.
@no_stp_on_snek: This is huge. 3090 class gpus rejoice!
Unsloth AI releases quantized Qwen3.6 models that run 2.5× faster on consumer GPUs, with the 27B model fitting in 24GB VRAM and the 35B-A3B achieving high throughput.