interesting aspect here is that when Liang talks about talent and DeepSeek being "just ordinary people, not geniuses", he doesn't mean IQ or gaokao or even…
RT Nat GurlainBackup of original transcript (Liang Wenfeng DeepSeek investors call)- https://archive.ph/NLuG9Teortaxes▶️ (DeepSeek 推特🐋铁粉 2023 – ∞): Full transcript of Wenfeng's investor conference call. He's a…
A Chinese article compiled 52 remarks from Liang Wenfeng’s four-hour investor meeting. I’ve summarised the most important ones below. DeepSeek has one central objective: AGI. This…
a quiet day lets us highlight a new neolab win.
Ironically, this means that DeepSeek's compute spend though 2026 will be covered 100% by Liang's personal investment. He also isn't buying data. The «funding round» is…
RT Teortaxes▶️ (DeepSeek 推特🐋铁粉 2023 – ∞)Re Liang Wenfeng believes that the comprehensive gap in AI between China and the US is 12-18 months, just like…
RT maxRunescape Bench's pareto-optimal lineup is super diverse.Fable -> Terra -> Grok -> GLM -> Deepseek -> Laguna all holding a solid chunk of coastline
Benchmark Solar Open 2 250B-A15B Solar Open 100B 102B-A12B Command A+ 218B-A25B Mistral Medium 3.5 128B dense, high MiMo-V2.5 310B-A15B DeepSeek-V4-Flash 284B-A13B, max Know. & Reasoning…
when DeepSeek will fill out the "Accelerator Hardware" quadrant, it'll be so gloriousbtw, what? Alibaba has its own chipsmaria: let this be a reminder that you…
US tech will keep benchmarking on DeepSeek until morale improves and some other lab with a noteworthy open model discloses its internal economics like this. One…
Model Size Terminal-Bench 2.1 SWE-bench Multilingual SWE-Bench Pro (Public Dataset) DeepSWE SWE Atlas (Codebase QnA) Toolathlon Verified Laguna S 2.1 118B-A8B 70.2% 78.5% 59.4% 40.4% 46.2%…
how are bong journalists real? Kimi is now responsible for the selloff too? This is dumber than the original DeepSeek momentCarl Zha: Hell Yeah! About time!
https://np.reddit.com/r/DeepSeek/s/skO7urrE2C DS4 sort of came and went from the spotlight. The consensus seemed to be that its most notable feature is its price, and then we…
«Industry observers are already calling it a potential “DeepSeek moment” for the global energy storage industry.»lmaowell, sodium will be a shock to someKyle Chan: CATL already…
RT Teortaxes▶️ (DeepSeek 推特🐋铁粉 2023 – ∞)Re @AcerFur @KuterDinel Funny reasoning from Kimi
about to write a bear scenario for DeepSeek and won't even get a few millions from Dario in return. pure love of truth
A single 24GB GPU is the practical floor for serious local inference. This guide compares six open-weight models that fit one card at Q4_K_M. It covers…
DeepSeek staff brushes off accusations of Fable routing永雏塔菲: @teortaxesTex @ilmalfalcon I really don’t believe ds would do this. (the two in the pic are ds employees.)
> when will DeepSeek harness be released?> the plan is to release it together with the official version [of V4]this makes a LOT more sense than…
RT Teortaxes▶️ (DeepSeek 推特🐋铁粉 2023 – ∞)Re My specific reason to think that Lisan's projection for the Chinese trend is wrong (irrespective of the gap convergence/divergence…
I am using unsloth Q8 Deepseek v4 Flash. So far I am able to run properly with the following command CUDA_DEVICE_ORDER=PCI_BUS_ID CUDA_VISIBLE_DEVICES=0,2,1 llamacpp/llama.cpp/build/bin/llama-server --model unsloth/DeepSeek-V4-Flash-GGUF/UD-Q8_K_XL/DeepSeek-V4-Flash-UD-Q8_K_XL-00001-of-00005.gguf --port…
Three open MoE flagships face off on measured intelligence, MIT versus Modified MIT weights, and real serving cost The post Kimi K3 vs DeepSeek V4 Pro…
I think that at this point DeepSeek can basically afford to flail and stay mediocre for the entire V4 generation (not that I wish for it,…
Deepseek V4 is about to be released; they say it will be cheap, and from the videos I've seen, it will be similar to Kimi K3…
DeepSeek shocked the field in late 2024 with DeepSeek-V3, then again with R1, a reasoning model trained at a fraction of the budget Western labs spend. Current lines: V3.2-Exp, R1, V3. The most-talked-about Chinese AI lab of the cycle.
Owner: DeepSeek. We have 280 stories indexed for this model, auto-tagged from titles across every tracked source — official announcements, papers, GitHub release notes, and third-party press. The CTA on each card links to the original; the official site is www.deepseek.com.
Related text models: GPT, Claude, Gemini, Gemma, Llama, Mistral, Grok, Qwen.