Trouble with DeepSeek flash

LLM apps3 posts from 3 people3d active+00 posts in the last 7 days, 2 the 7 days before (fading)

Posts per day3 posts, Sep 1 to Sep 25

The posts behind it

3, newest first

Post	Where	Date
[RFC]: Asymmetric P/D Deployment for DeepSeek-V4.1 Flash## Summary. DeepSeek-V4.1's bounded-replay structure creates an opportunity to make Prefill/Decode disaggregation substantially more asymmetric. Today, a Prefill worker still loads the full 40-layer model even though, after layer 20, only the final 128 prompt tokens need to be...vllm-project/vllmgz944367214	vllm-project/vllmgz9443672142 reactions, 5 replies	Sep 178 days ago
[Bug]: Garbled output for GLM 5.3 Flash when weights loaded with run:ai distributed streamer### Your current environment The output of python collect_env.py ### 🐛 Describe the bug Serving GLM-5.3-Flash with `--load-format runai_streamer --model-loader-extra-config '{"distributed": true, "concurrency": 16}'` and streaming weights from S3 produces intermittently...vllm-project/vllmMadhavJivrajani	vllm-project/vllmMadhavJivrajani2 reactions, 3 replies	Sep 1213 days ago
[Cloud] DeepSeek V4.1 Flash requestWould be very cool if we could have this model at disposal: https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash-Vision-Expollama/ollamamatbgn	ollama/ollamamatbgn3 reactions, 3 replies	Sep 13 weeks ago

Companies and products named

Company or product	Posts naming it
DeepSeek	3
vLLM	2
Hugging Face	2
NVIDIA	2
GitHub	1
Ollama	1

About this problem

Evidence

3 posts from 3 people in 2 places, about 1 a week over 17 days. Mostly on vllm-project/vllm, ollama/ollama. Tools named alongside: vLLM, Hugging Face, NVIDIA, GitHub.

Frustration Frustration 0 of 3· Seen on GitHub issues

How it was grouped

Posts about DeepSeek with no known issue, sharing the word "flash". First post Sep 1, 2026, latest Sep 17, 2026. Corroborated: 3 or more posts from 2 or more people or places. Method

Posts per day

The posts behind it

Companies and products named

Rising problems by email