HyperCLOVAX-SEED-Think-32B
Where it runs 8K context · ≥ 5 t/s floor
| Setup | Best plan | Memory | Generate |
|---|---|---|---|
| 8 GB GPU + 32 GB RAM | Q4_K_Mllama.cpp CUDAGPU + CPU layers · 21/72 layers on GPU | Fits | ~1.2–1.4 t/s Estimated extrapolated |
| 12 GB GPU + 32 GB RAM | Q4_K_Mllama.cpp CUDAGPU + CPU layers · 34/72 layers on GPU | Fits | ~0.9–1.1 t/s Estimated extrapolated |
| 16 GB GPU + 64 GB RAM | Q4_K_Mllama.cpp CUDAGPU + CPU layers · 47/72 layers on GPU | Fits | ~2.4–2.9 t/s Estimated extrapolated |
| 24 GB GPU + 64 GB RAM | Q4_K_Mllama.cpp CUDAFull GPU | Fits | ~26–32 t/s Estimated |
| 32 GB GPU + 128 GB RAM | Q4_K_Mllama.cpp CUDAFull GPU | Fits | ~40–64 t/s Estimated |
| 2× 24 GB GPUs + 128 GB RAM | Q4_K_Mllama.cpp CUDAFull GPU | Fits | ~23–30 t/s Estimated |
| 96 GB workstation GPU + 256 GB RAM | Q4_K_Mllama.cpp CUDAFull GPU | Fits | ~41–65 t/s Estimated |
| Mac, 24 GB | Does not fit | ||
| Mac, 48 GB | Q4_K_Mllama.cpp MetalUnified memory | Fits | ~8.6–10 t/s Estimated |
| Mac, 128 GB | Q4_K_Mllama.cpp MetalUnified memory | Fits | ~14–21 t/s Estimated |
| Mac, 512 GB | Q4_K_Mllama.cpp MetalUnified memory | Fits | ~16–24 t/s Estimated |
| DGX Spark 128 GB | Q4_K_Mllama.cpp CUDAUnified memory | Fits | ~9.2–10 t/s Estimated |
| Ryzen AI Max+ 395, 128 GB | Q4_K_Mllama.cpp VulkanUnified memory | Fits | ~6.2–9.0 t/s Estimated |
| CPU only, 64 GB DDR5-5600 dual channel | Q4_K_Mllama.cpp CPUCPU only | Fits | ~0.9–1.1 t/s Estimated extrapolated |
Same plan engine as the planner, with its default reserves and RAM. How plans are chosen.
Files exact Hugging Face bytes
| Artifact | Size | Files | Revision | Source |
|---|---|---|---|---|
| Q4_K_M naver-ellm/HyperCLOVAX-SEED-Think-32B-GGUF | 19.88 GiB | 1 | de2e200fce | HF files |
| EXL3 4-bit turboderp/HyperCLOVAX-SEED-Think-32B-exl3 | 17.60 GiB | 3 | 395babc505 | HF files |
| Native safetensors BF16 naver-hyperclovax/HyperCLOVAX-SEED-Think-32B | 62.05 GiB | 14 | 3c2b633861 | HF files |
Not found in the representative quant search (not proven unavailable): GGUF FP16, GGUF Q8_0, GGUF Q5_K_M, MLX 8-bit, MLX 4-bit, AWQ, GPTQ. Complete selected file sets; projectors, MTP/draft and lookup files are listed separately and budgeted only by plans that use them. File bytes are not allocated memory.
Quality
| AA evaluated mode | Index | LiveCodeBench | Terminal-Bench 2.1 | GPQA | HLE | Source |
|---|---|---|---|---|---|---|
| HyperCLOVA X SEED Think (32B) AA estimate | 11.4 | 13.7% | — | 61.5% | 5.5% | source |
Arena text rating
Unknown
Publisher-reported benchmarks
None extracted from the model card.
AA and Arena describe the hosted model in the stated mode, not each quantized file; scores are never inherited by base, pruned, abliterated or quantized variants and never averaged together.
How to run it
No model-specific recipe in the research. The planner binds the backend’s documented command template (llama.cpp, Ollama, LM Studio, MLX, vLLM) to your plan: open a plan’s details.
Recipes are source examples (settings are the publisher’s or author’s, not measured optima) and were not executed by LocalFit. A downloadable file does not prove your backend build supports the architecture.
Reported speeds 0
No reported measurement yet. Submit one.