DeepSeek V4 Flash on a Single AMD MI300X(github.com) |
DeepSeek V4 Flash on a Single AMD MI300X(github.com) |
Edit: Now I think about it, this might be the cheapest way to run the DeepSeek V4 Flash 0731 on a dedicated inference server at original weights. I haven’t run mixed load benchmarks but I guess it’s possible to generate $3-$4 worth of tokens per hour and still maintain a usable per-user throughput.
I found an offer to rent two at $1.65 per hour https://spark.enverge.ai/#pricing
The MI300X will vastly outperform it for only a slightly higher price.
Otherwise, you can probably buy one of these second hand from somewhere (SXM A100s are available that way) and run it in an adapter board.
The only solution is more companies making high end units, only competition will make it better for consumers.
Luckily, DeepSeek V4 Flash will run in 144GB too because it's 256 MoE exports are native MXFP4 quantized.
When they just loaded the weights alone, it was taking 156GB in vLLM. After warm-up and adding a KV cache pool, it took over 200GB.
And this implementation is already cutting down the 1M token context window you would normally get.
That being said, if you're dropping tens of thousands of dollars on graphics cards then picking up a rackmount case to go around the card is pretty insignificant.
I assume this is parallel work.
we did some work on this for 2xMi300x (kindly referenced in the readme) https://blog.doubleword.ai/deepseek-v4-flash-mi300x. https://hotaisle.xyz/quick-start hotaisle is great for getting Mi300x to experiment with
Dumbed down quantization?
No. Full intended inference weights preserved, so far so good.
Slow performance?
No again. Looks like you could get over 150 tokens/second.
Give up context window size?
Yes. Original model is trained for and served at 1M, this is 256k. A very practical tradeoff though. Codex is in this range, and quality does start to drop off toward the full size.
Was also surprised to learn just how much of Codex's window was being burnt on shit I didn't want or use. Sure I can pass this and that flag to eliminate most of it, but for a $200/mo product aimed at professionals, that isn't something anyone should have to janitor (also totally ignoring the bandaid of banked resets they've slapped over their repeated mistakes).
It's wild just how far $20 will get you with Deepseek, even at their new rates. Buyers Remorse is my very least favorite feeling, I felt sick thinking about what the $1200 I had given OAI this year would have gotten me had I only tried sooner.
also it would be interesting to figure from the DSpark paper whether their numbers are consistent with the GPUs still being H800s, since they never actually say...
BTW, how many many params open weight frontier models have? A few teras, 100s of teras?
Qwen3.8-Max: 2.4T
DeepSeek V4 Pro: 1.6T
DeepSeek V4 Flash: 284B
(all are total parameter counts, not active parameters)
Like linux DMABUFs with userland hardware command ring buffers (I guess this hardware ring buffer instance would be specific to a VMID and a PASID).
I haven't seen the cards all by themselves yet.
(274gb/s vs 5.3tb/s)
Also if they are keeping the price low because they want to gain market share and reduce the competitiveness of Chinese models they won't be able to raise prices without providers serving open models (at cost + low margin) severely undercutting them.
No bubble, where money?
Not exactly super special, but a ~constant 600W+ of heat tends to be a learning experience. It's worse than a high end gaming rig, much closer to a literal space heater. I don't work during the summer because it sucks fighting both this and the sun with AC. I do have a fan that slots into the window and can push or pull, but kicking the waste heat outside doesn't help when its humid.
I can tell you it will pop in 10 years and when it pops, it will still be 20x bigger than in 2026. Does that even make any sense?
People said AI bubble will pop soon in 2024 and that it was overvalued. Turns out, many AI stocks 10x, 20x since 2024. Actual usage has gone exponential as well. Anthropic revenue went from $100m ARR at start of 2024 to $80b ARR today.
Well, given the literal trillions being spent, the only ways this pays off are:
1. AI replaces a non-trivial fraction of human employees.
2. Someone builds a Culture Mind, and humans become (hopefully) pampered pets of AIs we don't understand. Seems unlikely, but it would arguably count as a payoff even if it made money meaningless.
Or maybe the AIs don't want pets, and you get SkyNet. Which definitely doesn't care about paying off anyone's investments.
When you look at various news articles about investors, yeah, there are definitely a lot of rich people who think that they're going to automate all human labor or just bring about the Singularity. Possibly with them in charge of the rest of us. If you don't make these kinds of wild assumptions, then yeah, this is looking like one of the biggest bubbles ever.
Debt-backed investors have to pay up eventually. =3
My personal take, with no sources: 100T sounds excessively high given they need to be able to actually serve these things on commercially available hardware. I would guess they are in the same order of magnitude as the Chinese frontier models. It's possible their edge is in RL training methods, training-time compute, and access to data (e.g. from customers' CC/Codex sessions), not in model size.
Interest alone, at assumed 5%, amounts to about $150 billions per year. That's probably higher than the combined AI revenue of the top 3 providers.
Now let's build the model out more. What is the projected revenue, backlog, improvements in existing big tech businesses such as AI helping Meta's ad business?
I mean, it's pretty clear that's going to happen. How could it not?
The only question is whether resources are optimally allocated at the moment to prepare for this. That seems unlikely at best. So yes, there is probably a bubble, and if so, then yes, it will pop, and then life will go on, with resources better allocated. Just like when the dot-com bubble popped.
This is so hard to sort the true from the false nowadays.
If LLM becomes that good at coding, I'll have to run an open weight one locally.
To be clear this project doesn’t aim to achieve the best inference economics per token. MI300X doesn’t have native MXFP4 so it’s not even the right platform for the model. That’s why very few deployment recipes are available.
It’s interesting to me because MI300X is quite accessible to a small team with budget for just 1-2 GPUs. DeepSeek V4 Flash otherwise wouldn’t even fit on 2x H100s.
We can run several coding agents during the day and batch inference jobs overnight and serve the entire team with guaranteed privacy, without compromising precision or speed.
In fact we found that many inference providers are quantising the weights or even KV cache, and due to the low prices they serve at massive batches, resulting in unstable throughput. I ran GSM8K as a quick validation test and this deployment is “better” than the OpenRouter endpoint in a statistically significant way (I wouldn’t name the provider here). I will run some follow up benchmarks and update the repo when I find some time.
You shouldn't rent one out if you're just serving it for yourself, but from a financial standpoint if you sell to users you can take a 100% margin.
This is not the real margins, even if you are selling to 8 users it's 90 tps per median stream. So assuming that .6-.7$
This is not even remotely worth it.
You need to 3x this tps(~1500 tps) to be worth it, and that's what most providers are doing, at 20-30 users at 50-60 tps with better optimized batch processing and kernels you can make some profit.
Lets assume that you've got 2 million input tokens, 1 million output tokens and 98.5 million cached tokens to process. That would cost 2 * $0.14 + 1 * $0.28 + 98.5 * $0.0028 = $0.8358 with DeepSeek API pricing.
For comparison, it would take 2M / 8000 + 1M / 800 = 1500 seconds to process this amount of tokens with the linked framework, which is about $0.83 when we assume $2/hr for one MI300X.
However, other inference providers have 10 times higher prices for cached tokens, which results in a comfortable margin.
And we should not discount that DeepSeek also gets paid in data, which is probably more valuable to them.
And I believe that this framework still has some room for optimization for generation with high batch sizes.
Based on your $0.83 estimate, the margin isn't great. This is within shooting distance of "at cost" which is probably pretty close to what DeepSeek is operating with, ignoring the value of the data they're collecting of course.
> And I believe that this framework still has some room for optimization for generation with high batch sizes.
If that optimization can bring this scenario closer to $0.50 then it gets pretty compelling, otherwise I'm not confident.
That's agentic feedback loops for training, right? Any more detail on this, such as how they actually tell whether that data is good or not? That seems like a very hard problem, and like the value of that data is low compared to just building their own, controlled RL gyms.
Definitely not. Inference is not as expensive to operate as many people seem to assume. The frontier labs are probably making a lot of money from selling tokens. It’s covering all of the R&D costs like salaries, collecting training material, and running the large training operations that costs a lot of money.
Are you claiming that the frontier labs like OpenAI and Anthropic are actually making a profit contrary to all the claims?
Serving local models has advantages other than price. If you work in restricted industries, or have a strong need to protect your IP, or if you just value privacy more than cost, you now have options.
As others have mentioned, there's the privacy factor as well.
I think the margins of DeepSeek may be a bit better than with this vibe-coded framework here, since they had the liberty of optimizing their models for their own hardware.
For DeepSeek V3, they claimed a cost profit margin of 545%: https://github.com/deepseek-ai/open-infra-index/blob/main/20...
At the time, open frameworks were not anywhere close to achieving that number. Not sure whether they caught up. The software wizards at DeepSeek are quite skilled.
I am not sure how they are using the data for post-training, but there probably are ways to get signal out of it, e.g. sentiment analysis when the user begins cursing at the agent, or checking whether the user continued another session with the generated code, or started a new session with the same starting point as before, i.e. they git-stashed.
Generally, you can train on data that is quite bad (e.g. the entire internet). It will still work, but take much longer compared to clean data.
Thank you for elaborating, that's already useful. Anywhere I can learn more about this? I'm very interested in it!
For more LLM-specific stuff, you can pick some agent trace dataset on https://huggingface.co/datasets?format=format%3Aagent-traces and check out what people are doing with it (usually linked on the right when you click on a dataset).
And of course https://scholar.google.com/ for research papers.