So can a lot of other models.
“This load can be towed by a single vehicle”
DeepSeek's distilled new R1 AI model can run on a single GPU | TechCrunch
Submitted 10 months ago by schizoidman@lemm.ee to technology@lemmy.world
https://techcrunch.com/2025/05/29/deepseeks-distilled-new-r1-ai-model-can-run-on-a-single-gpu/
Comments
LodeMike@lemmy.today 10 months ago
blarth@thelemmy.club 10 months ago
7b trash model?
LainTrain@lemmy.dbzer0.com 10 months ago
I’m genuinely curious what you do that a 7b model is “trash” to you? Like yeah sure a gippity now tends to beat out a mistral 7b but I’m pretty happy with my mistral most of the time if I ever even need ai at all.
knighthawk0811@lemmy.world 10 months ago
it’s distilled so it’s going to be smaller than any non distilled of the same quality
TropicalDingdong@lemmy.world 10 months ago
Yeah idk. I did some work with deepseek early on. I wasn’t impressed.
HOWEVER…
Some other things they’ve developed like deepsite, holy shit impressive.
double_quack@lemm.ee 10 months ago
Save me the search, please. What’s deepsite?
vhstape@lemmy.sdf.org 10 months ago
the Chinese AI lab also released a smaller, “distilled” version of its new R1, DeepSeek-R1-0528-Qwen3-8B, that DeepSeek claims beats comparably sized models on certain benchmarks
Most models come in 1B, 7-8B, 12-14B, and 27+B parameter variants. According to the docs, they benchmarked the 8B model using an NVIDIA H20 (96 GB VRAM) and got between 144-1198 tokens/sec. Most consumer GPUs probably aren’t going to be able to keep up with
brucethemoose@lemmy.world 10 months ago
Depends on the quantization.
7B is small enough to run it in FP8 or a Marlin quant with SGLang/VLLM/TensorRT, so you can probably get very close to the H20 on a 3090 or 4090.
avidamoeba@lemmy.ca 10 months ago
It proved sqrt(2) irrational with 40tps on a 3090 here. The 32b R1 did it with 32tps but it thought a lot longer.
fogetaboutit@programming.dev 10 months ago
ew probably still censored.
T156@lemmy.world 10 months ago
The censorship only exists on the version they host, which is fair enough. If they’re running it themselves in China, they can’t just break the law.
If you run it yourself, the censorship isn’t there.
jaschen@lemm.ee 10 months ago
Untrue, I downloaded the vanilla version and it’s hardcoded in.
MonkderVierte@lemmy.ml 10 months ago
Yeah, i think the censoring in the LLM data itself would be pretty vulnerable.
Mwa@lemm.ee 10 months ago
You can self host it right??
jaschen@lemm.ee 10 months ago
The self hosted model has hard coded censored content.
fogetaboutit@programming.dev 10 months ago
if the model is censored… then what, retraining it? Or doing it from scratch like what open-r1 is doing?