Nicheloom visits: 35,311 all-time · 10,180 in the last 7 days

The opportunity tracker for new startups.

Today: 1,436 launches · Niche of the day: browser-based utility tools · See what launched today →

DeepSeek-V4.1-Flash-1017-FP8-abliterated-4x-RTX-PRO-6000

Abliterated DeepSeek-V4.1-Flash-1017 (FP8 dense, FP4 experts) on 4x RTX PRO 6000 with vLLM-Moet r21 and DSpark: 371-376 tok/s single-stream agent decode at 30k-120k context, 1M context, Flex-start GCE VM

Get picks like this daily. Today's launch numbers, the top launches, AI and tech news, and a daily opportunity pick, straight to your inbox.

Visit website ↗ Source listing ↗

GITHUB · launched Oct 23, 2026 · 27 upvotes (top 48%)

Similar launches

See all 252 similar launches →

Could you build this?

No Running a 1M-context abliterated DeepSeek-class model in FP8 at 166-211 tok/s on four RTX PRO 6000 GPUs needs custom inference kernels, model surgery, and multi-GPU serving expertise, plus expensive hardware.

What it would actually take: A real version needs a multi-GPU inference stack built on vLLM or a similar engine, with custom FP8 kernels, tensor and pipeline parallelism, and speculative decoding such as the 'DSpark' component. The model would need abliteration, which means editing refusal-direction weights and validating the result, and then rigorous benchmarking. This requires ML systems engineers and a server with roughly 384 GB of VRAM.

Competitors

Other products that read as similar to this one : 252 launches clear the similarity bar, closest 8 shown.

Attention rank: #136 of 253 (itself plus its competitors, highest first, normalized so YC and Product Hunt are compared fairly).

Launched 202 days after the earliest competitor.

Other launches for this product