# basetenlabs/deepseek-ultrafast

> A high-performance Rust inference engine for DeepSeek-V4.1-Flash on sm_100a/sm_103a GPUs, supporting 1M context and multi-server KV cache sharing.

- Magnitude: 5.9 out of 10 — Early signal
- Stars: 170 total · +58 stars measured 2026-11-14, 00:49–10:32 UTC, ≈ 122 by evening
- Star trust: star growth looks organic
- Category: Language models · Language: Rust · Created: 2026-10-27 · Last push: 2026-11-14
- GitHub: https://github.com/basetenlabs/deepseek-ultrafast · Page: https://gitnova.dev/en/r/basetenlabs/deepseek-ultrafast

## Useful for

- Run a local OpenAI-compatible DeepSeek-V4.1-Flash server on 4 or 8 GPUs
- Deploy a multi-server setup with shared KV cache via Redis and RDMA
- Benchmark decode speed and acceptance length on your own hardware

## Why it’s here

- Star-counter measurements on 2026-11-14 (UTC), 00:49–10:32: 112 → 170 stars (+58). This is the change over that interval.
- Estimated end-of-day forecast: about +122 stars, using observed gains and the previous day.
- Before this, the repository barely got any stars — about 0 per day.
- Elevated gains were measured on the last 2 completed UTC days in a row. Continued growth is not guaranteed.
- The repository is 18 days old and already has 170 stars.
- Hacker News: “DeepSeek Ultrafast” — 53 points, 17 h ago.

## Star trust

Star growth looks organic.

Star-trust labels are heuristics based on the repository’s behavior, not a check of every stargazer.

## Numbers

- Forks: 4
- Issues and pull requests: 1
- Watchers: 0
- Average over the last week: 40 per day
- Usual pace: 0 per day
- Stars in the last hour (measured): 3

## Stars per day, last 19 days (oldest → newest, today is partial)

2026-10-27 … 2026-11-14: 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 57, 103, 58

## Hacker News

- DeepSeek Ultrafast — 53 points, 10 comments: https://news.ycombinator.com/item?id=50486914

## Spotted in now

- Spotted on Hacker News

## Similar by description

1. **mw00/project-maya** — 2.1 · Steady · Language models · C++ · +11 stars measured 2026-11-14, 00:07–10:36 UTC, ≈ 28 by evening
   Local engine, server and dashboard to run the 321B MoE model GLM-5.3-Flash on one or two NVIDIA GPUs, offloading experts to RAM and NVMe.
   Full card: https://gitnova.dev/en/r/mw00/project-maya.md
2. **FlashML-org/FreeToken** — 2.1 · Cooling · Language models · Python · +15 stars measured 2026-11-14, 00:06–10:36 UTC, ≈ 39 by evening
   A Mixture-of-Experts inference engine for running frontier-scale open-weight MoE models (290B+) on consumer hardware — GPUs, CPUs, and host memory. Aimed at users who want to run frontier models locally.
   Full card: https://gitnova.dev/en/r/FlashML-org/FreeToken.md
3. **deepseek-ai/3FS** — 1.0 · Steady · DevOps & infrastructure · C++ · +1 star measured 2026-11-14, 00:12–10:39 UTC, ≈ 3 by evening
   3FS is a distributed file system from DeepSeek for AI training and inference workloads, using SSDs and RDMA networks as a shared storage layer.
   Full card: https://gitnova.dev/en/r/deepseek-ai/3FS.md
4. **Neroued/ninfer** — 0.7 · Steady · Language models · C++ · +0 stars measured 2026-11-14, 00:12–10:40 UTC, ≈ 1 by evening
   NInfer is a from-scratch C++/CUDA inference engine for Qwen3.5 Dense and MoE models on a single NVIDIA RTX 5090. It serves text, image, and video prompts via a local CLI or OpenAI-/Anthropic-compatible HTTP APIs.
   Full card: https://gitnova.dev/en/r/Neroued/ninfer.md
5. **NVIDIA/TensorRT-LLM** — 1.3 · Steady · Language models · Python · +3 stars measured 2026-11-14, 00:12–10:39 UTC, ≈ 6 by evening
   NVIDIA library for optimizing inference of large language models and visual generative models on GPUs, with a Python API, specialized kernels, and an efficient C++/Python runtime.
   Full card: https://gitnova.dev/en/r/NVIDIA/TensorRT-LLM.md

---

Magnitude (0–10) measures how fast and how unusually interest in a repository is growing right now. It is not a quality score.
Days are UTC. “So far today” is a fact; “expected by the end of the day” is a forecast.
Summaries and use cases are written by an LLM (DeepSeek V4.1 Flash) from the README and may be inaccurate: verify specific claims (benchmarks, speed, hardware) in the repository itself.
Data as of 2026-11-14 10:41 UTC, updated every 30 minutes.