@TheAhmadOsman: To clarify, OMP has 97.5% cache hits Most of my usage is self-hosted models, and not all inference setups report the to…
Summary
A user clarifies that OMP has a 97.5% cache hit rate and discusses the use of self-hosted DeepSeek V4 Flash models, addressing concerns about caching performance in inference setups.
View Cached Full Text
Cached at: 08/23/26, 11:43 PM
To clarify, OMP has 97.5% cache hits
Most of my usage is self-hosted models, and not all inference setups report the token telemetries properly so that’s why it was missing in my original screenshot
These are DeepSeek V4 Flash 0731 running through OMP in the last couple of days https://t.co/BC5EL5Sxet
Theo - t3.gg (@theo): Holy shit is OMP actually that bad at caching???
Similar Articles
@TheAhmadOsman: DeepSeek Harness is so good 99% cache hits is crazzzzyyyyy
The tweet praises DeepSeek Harness for achieving an impressive 99% cache hit rate, highlighting its efficiency in system performance.
I tested DeepSeek Harness with GLM, Kimi, Opus, and GPT to see if prompt caching still works with other models
The article tests whether DeepSeek Harness maintains high prompt caching rates when using alternative AI models, finding that GLM and Kimi achieve 97-99% cache reuse, while Opus shows no cache activity and GPT test failed.
DeepSeek-v4.1 Flash: Pushing the Limits of KV Cache Compression
This article analyzes the DeepSeek-V4.1 Flash model, detailing its technical report on KV cache compression and architectural optimizations that enable efficient long-context processing and high-speed inference.
@ciruai: Testing DeepSeek v4 Flash on the AMD Ryzen AI Max+ 395 Strix Halo with 128GB RAM. Getting ~15 TPS over a decently long …
Testing DeepSeek v4 Flash on the AMD Ryzen AI Max+ 395 with 128GB RAM achieves ~15 TPS for a 284B MoE model (13B active) locally, costing $3,000 versus $25,000+ for a datacenter setup, highlighting the feasibility of running large models on consumer hardware.
@_avichawla: DeepSeek did it again! They just open-sourced DeepInfer, the inference stack behind cache hits that return the first to…
DeepSeek open-sourced DeepInfer, its inference stack that reuses KV cache from SSDs via 3FS to deliver first tokens up to 26x faster, enabled by MLA's 93% KV cache compression.