6 ms·
DeepSeek-v3.2-Exp
- terespuwash 1y agoLooks like Deep Sparse Attention can help with code (structured and long-file reasoning)
- matrix2596 1y agoawesome that sparse attention used in real world setting
- mythz 1y agoHappy to see Chinese OSS models keep getting better and cheaper. It also comes with a 50% API price drop for an already cheap model, now at: $0.28/M Input ($0.028/M cache hit) > $0.42/M Output
- manishsharan 1y agoThis price drop is nice but I wonder how long it will last. Their prices used to be very low,then they almost doubled, and now it dropped.
- nacs 1y agoI don't know if it will stay this low but the whole point of v3.2 is to be cheaper to run than <= v3.1. (The inference costs are cheaper for them now as context grows because of the Sparse attention mechanism)
- guluarte 1y agoI was using it daily, but after the price jump, using codex and claude was much cheaper than using deepseek.
- dizhn 1y agoWhat was the price before? I thought they had just increased their prices.
- espadrine 1y agoInput: $0.07 (cached), $0.56 (cache miss) Output: $1.68 per million tokens. https://api-docs.deepseek.com/news/news250929 https://api-docs.deepseek.com/news/news250929
- Havoc 1y agowow...gigantic reduction in cost while holding the benchmarks mostly steady. Impressive.
- techlatest_net 1y ago[dead]
- awongh 1y agoThe 2nd order effect that not a lot of people talk about is price: the fact that model scaling at this pace also correlates with price is amazing. I think this is just as important to distribution of AI as model intelligence is. AFAIK there are no fundamental "laws" that prevent price from continuing to fall, at least correlated with Moore's law (or whatever the current AI/Nvidia chip development cycle is called right now)- each new generation of hardware is significantly faster/cheaper than the next- so will we see a ChatGPT-5 model at half the price in a year? (yes I know that thinking models cost more, but just on a per-token basis)
- samuelknight 1y agoYou are vastly underestimating the price decline. To cherrypick one article; in the first two years since GPT 3.5, inference price for the same amount of intelligence has decreased 10x per year according to a study by Andreessen Horowitz https://a16z.com/llmflation-llm-inference-cost/ https://a16z.com/llmflation-llm-inference-cost/. So in a stark slowdown scenario, we could still see a 1000x decrease in the next 5 years. Price deflation is not tied to Moore's right now because much of the performance gains are from model optimization, high bandwidth memory supply chains, and electrical capacity build out, not FLOP density.
- awongh 1y agoTrue! I just know that model optimization gains are much less guaranteed than say, FLOP density, even though model optimization has so far provided way more gains than hardware advancements. Part of me is optimistic that when the AI bubble bursts the excess data center capacity is going to be another force driving the cost of inference down.
- NemoNobody 1y agoHaha, I love how delusional everyone is about AI. Yeppers, when that bubble burst - that's hilarious. This is the kinda stuff grandkids won't believe someday.
- naasking 1y ago
- wwizo 1y agoYou guys rock! I'm very curious how will this perform against real word data, where small nuance matters. Also have you tested it beyond 128K context window?
- esafak 1y agohttps://openrouter.ai/deepseek/deepseek-v3.2-exp https://openrouter.ai/deepseek/deepseek-v3.2-exp
- nacs 1y agoStrange - the model is marked as "Trains on data" ("To our knowledge, this provider may use your prompts and completions to train new models. This provider is disabled, but it can be re-enabled by changing your data policy."). This is usually not the case for paid models -- is Openrouter just marking this model incorrectly or do Deepseek actually train on submitted data?
- esafak 1y agohttps://cdn.deepseek.com/policies/en-US/deepseek-privacy-policy.html https://cdn.deepseek.com/policies/en-US/deepseek-privacy-pol... https://openrouter.ai/docs/features/privacy-and-logging#data-retention--logging https://openrouter.ai/docs/features/privacy-and-logging#data... It seems so.
- seunosewa 1y agoIt is no longer the case that paid providers don't train on your data on Openrouter. You can exclude such sources in the settings.
- nacs 1y agoYep I have that setting disabled so the number of providers for that model on Openrouter currently is 0 for me. I guess I'll wait for a 3rd party provider on Openrouter that doesn't log DS 3.2.
- echelon 1y agoIs Open Router really open? I see their "main" repo as archived and various smaller projects. Is it just the API client bindings that are open and the core routing service is closed!
- 1y ago
- mmastrac 1y agoInteresting that models still evolve fast enough that dedicated model-specific hardware isn't a big contender right now. We're still seeing major scaling gains on mostly generic platforms.
- gunalx 1y agogoogle tpm, groq and cerebras needs yo be mentioned even if they are more general architecture optimized.
- ramshanker 1y agoWhat happened to Meta Open weights models? Lately I keep hearing more of Deepseek than LAAMA?
- Alifatisk 1y agoWasn't The Llama 4 maverick and scout a flop?
- grim_io 1y agoOne huge problem with these "cheap" models is that they happen to be more expensive in the typical agent workflow if the provider does not support caching. Input and output costs are peanuts compared to the order of magnitude(or more) amount of tokens that hit the cache. At that point you might as well use GPT-5. It will be the same price or cheaper, and more capable.
- NotMichaelBay 1y agoI was under the impression that this model does support caching. The pricing page says the cost of input tokens (cache hit) is $0.028.
- segmondy 1y agoyou declared a huge problem and followed up with an IF. deepseek API supports caching, stop manufacturing problems where there is none. https://api-docs.deepseek.com/guides/kv_cache https://api-docs.deepseek.com/guides/kv_cache
- grim_io 1y agoSure. But there is no way I'm going to use the deepseek endpoint. Openrouter says they might use your data for training.
- cheema33 1y agoFirst you complained about lack of caching. When you were informed that the model supports caching, instead of admitting your error you switched to an unrelated complaint. I hope that you you do not use similar strategies for discussion in your personal and work life.
- grim_io 1y agoYour broad attack on me as a person is unnecessary. If you read my post carefully, you will realize that I did not make any contradictory statements.
- eric15342335 1y agoNot sure if I get it correctly: They trained a thing to learn mimicking the full attention distribution but only filtering the top-k (k=2048) most important attention tokens so that when the context window increases, the compute does not go up linearly but constantly for the attention->[query,key] process (it does grow up linearly in the graph because you still need to roughly scan the entire context window (which an "indexer" will do), but just very roughly here in order to speed up things, which is O(L) here).
- impact_sy 1y agoPrices fall, benchmarks remain stable. Maybe in the future, LLM will spend most of its money on electricity.