Skip to content
#

deepseek-v4

Here are 365 public repositories matching this topic...

Project Maya - run GLM-5.3-Flash (321B MoE), DeepSeek-V4.1-Flash-1017 (748B MoE) and MiniMax-M3.1-Flash (428B MoE) privately on your own GPU(s), from one card to 16, NVIDIA or AMD, Linux or Windows. Engine, server, dashboard, OpenAI- and Anthropic-compatible API, quant tools.

  • Updated Oct 27, 2026
  • Cuda

Xolotl (formerly Strata-DSV4.1): run DeepSeek-V4.1-Flash, all 552B, as a local coding agent on one gaming PC (RTX 4090 + 90 GB RAM) at up to ~27 tok/s, 512K context, October 1017 builds included. Plus GLM-5.3-Flash and Qwen3.8. MoE experts tiered across VRAM, pinned RAM and NVMe; OpenAI + Anthropic API.

  • Updated Oct 28, 2026
  • C++
deepseek-local-guide-ru

🐋 Полный гайд на русском: как установить и запустить DeepSeek локально (V4, R1) на ПК, Mac и сервере — Ollama, llama.cpp, LM Studio, vLLM, Open WebUI, API, RAG, агенты, fine-tuning, лучшие фишки

  • Updated Oct 17, 2026
  • Shell

MCore-Bridge: Providing Megatron-Core model definitions for state-of-the-art large models and making Megatron training as simple as Transformers — with support for 300+ large language models (Qwen3-Next, GLM-5.3, Deepseek-V4, MiniMax-2.7, ...) and 200+ multimodal large models (Qwen3.8, Qwen3-Omni, Gemma4, ...).

  • Updated Oct 27, 2026
  • Python

玄机 XuanJi · Claude 客户端 + Claude Code Agent · 国内直连 Claude Sonnet 4.6 / Opus 4.7 / DeepSeek v4 · Claude 防封不退款, Claude Code 国产替代首选

  • Updated Jun 14, 2026
  • HTML

Add this topic to your repo

To associate your repository with the deepseek-v4 topic, visit your repo's landing page and select "manage topics."

Learn more