LocalClaw/LLM/DeepSeek V3.2 Exp (671B MoE)
Public model card reference

DeepSeek V3.2 Exp (671B MoE)

Catalogue summary: Experimental V3.2 with DeepSeek Sparse Attention (DSA) — halves inference cost vs V3.1 on long context while keeping quality. 128K context, improved coding & tool-use. MIT licensed. Server-grade.

Repository editorial metadata; verify comparative claims in the linked upstream material.

Public model cardNo public GGUF verified671B (37B active)2025-09
Choose an app

Only verified options can be opened. Unsupported apps are clearly marked.

Compare models
LM StudioNot available for this model
UnslothNot available for this model
OllamaNot available for this model
Open on Hugging FaceModel card, licence and access details
llama.cppNot available for this model
Use with LocalClawOptional workspace after the model is installed
Source status
Public model card
Public GGUF
Not verified
Parameters
671B (37B active)
Access
Public card

Verified source status

LocalClaw verified a public model card for DeepSeek V3.2 Exp (671B MoE) on 2026-09-07, but no public GGUF file in that repository. The catalogue RAM and quantization fields are estimates, not a verified install path.

Only the public model card was verified; no public GGUF file was verified in that repository. Open the model card to confirm current artefacts and supported runtimes. LocalClaw does not claim a one-click LM Studio install.

Public model card (no GGUF file verified 2026-09-07): deepseek-ai/DeepSeek-V3.2-Exp
chatcodereasoningquality

Source availability

01
Open the model cardUse the verified upstream card as the source of truth for this record.
02
Inspect current filesNo public GGUF file was verified in the repository during the catalogue audit.
03
Confirm a supported runtimeChoose a local runtime only after verifying the exact artefact, licence and hardware requirements upstream.

Catalogue record

  • Family: deepseek
  • Parameters: 671B (37B active)
  • Recommended quantization: Q4_K_M
  • Catalogue minimum RAM: 512 GB
  • Catalogue model size: 380 GB
  • Tags: chat, code, reasoning, quality

Practical limits

  • Catalogue RAM is a minimum estimate, not a guarantee for every context length or runtime.
  • Speed and memory use vary by quantization, backend, context length and system headroom.
  • Verify architecture, licence and usage restrictions in the linked upstream material before deployment.

Catalogue tags

  • chat
  • code
  • reasoning
  • quality

Capability profile

Repository catalogue ratings used by LocalClaw's editorial rubric. They are not a standardized third-party benchmark.

speed
2
quality
10
coding
10
reasoning
10

Technical notes

Developer
See upstream repository
License
See upstream repository
Context window
See upstream repository
Architecture
See upstream repository

Related catalogue entries

Linked mechanically by family, shared tags and nearby RAM tier; this is not a quality ranking.

DeepSeek R1 0528 (671B MoE) 671B (37B active)DeepSeek V4 Pro (1.6T MoE) 1.6T (49B active)DeepSeek R1 Distill (70B) 70BDeepSeek R1 Distill (32B) 32BDeepSeek R1 Distill (14B) 14BDeepSeek R1 Distill (8B) 8BDeepSeek Coder (33B) 33BDeepSeek Coder V2 (16B) 16B

Where to go next