Pinned Tweet
$NBIS is a winner. Can't you see it? Too many FinTwit thesis, wasted words. Jensen has SCREAMED it to you. He knew Arkady before we knew him.
If your smooth brain doesn't understand who Arkady is, I can't help you. Just focus, ask questions, model it, ask your favorite LLM, do what you do. It's in front of you. You need to decide.
Very excited to announce our new models: Nemotron 3.5 Super and Lightning!
This also feels like a good time to share that a few months ago I joined the amazing Nemotron Pretraining team. We're cooking some very large and very strong models for you!
Here is a tech summary for 3.5:
* Super is a 120B total / 12B active vision language model (VLM)
* Lightning is a 30B total / 3B active text-only LLM
* Super scores 51.78% on SWE-bench Pro (better than Nemotron 3 Ultra despite ~4.6× fewer active params!)
* arch: MoE Hybrid Mamba-Transformer designed for efficient inference, with a 1M token ctx window, and native speculative decoding (via MTP)
* training is done across our GB200 and GB300 (and deployment validated across all imaginable SKUs, including consumer (RTX) and DGX Spark)
A few details on how we train these:
-- Pretraining
* pipeline: initialize with Nemotron 3 base -> custom MTP initialization + 100B-token MTP warm-up -> 1T-token distillation from the next larger N3 teacher
* Lightning then goes through a 21B token long-context (LC) stage (@ 1M ctx)
* Super goes through:
- Indic CPT: 100B tokens at 8K (10B Indic + 90B replay, we built a synthetic corpus with 2.83T toks across 22 Indian langs! Indic benchmarks improve by up to 9.82%)
- LC extension: 35B tokens at 1M
- Agentic SWE mid-training: 65.4B tokens at 256K (uses whole repos, issue-PR pairs, commit histories, etc.); DeepSWE pass@1 improves from 6.4 -> 10.4
- LC recovery: 9B tokens at 1M (as we observed regression after mid-training)
-- Post-training
* Super: two-stage SFT -> joint RLVR/RLHF -> specialist teachers -> two-stage MOPD -> MTP boosting
* Lightning uses a lighter text-only pipeline: single-stage SFT -> RLVR
* Super's SFT transforms the text-only base into a VLM using vision, agentic, reasoning, chat, safety, LC data
* A few examples of RLVR environments:
- Olympiad-level competitive programming
- Finance agents using SEC search, document parsing, evidence retrieval, and tool use
* Specialist teachers include Agentic, Reasoning, Video reasoning and tracking, Vision tool-use, Vision-reasoning, and an Alignment/RLHF teacher
* Two-stage MOPD consolidates these specialist capabilities into one model
* Post-training ends with MTP boosting to better align the drafter and reduce recursive train–inference mismatch; this improves Super's SPEED-Bench acceptance length by 21.8% (to 4.612 tokens)
-- Inference
The model is highly optimized for inference (with plenty of quantized options to choose from)! A few data points:
* On a GB200, Super reaches 4,856 output toks/s/GPU with ≥50 toks/s/user on a 20-turn agentic workload
* Super provides 2.34x the throughput of Qwen3.5 122B-A10B and 1.90x the throughput of Qwen3.8 Flash-Next
* Native MTP is complemented by standalone parallel drafters: LiLiCorr for Super, and DFlash and DSpark for Lightning
Looking forward to seeing what you do with 3.5! Phase shift for Nemotron incoming! :)) (with the usual this will be our weakest model ever, but really really so)
paper: research.nvidia.com/labs/nem…
Now that Paramount Skydance’s acquisition of Warner Bros. Discovery has finally, mercifully closed, a look at the looming debt questions—and the broader cooling on media conglomerates since the Ellisons’ opening bid.
@WilliamCohan reports: puck.news/how-the-skydance-w…
H.C. Wainwright lowered the firm's price target on $UNCY to $13 from $15 and keeps a Buy rating on the shares after the company announced that it resubmitted its New Drug Application for oxylanthanum carbonate for hyperphosphatemia in patients with chronic kidney disease on dialysis with new Chemistry, Manufacturing, and Controls data from a new third-party drug product manufacturing vendor. The firm updated its model to reflect the NDA resubmission with the new vendor and assumes a six-month Class 2 resubmission review, so it now expects FDA approval at the end of Q1 of 2027, the analyst tells investors.
🗞️ There Will Be ‘Enormous Demand’ For Psychedelic Treatments, RFK Says, Praising Trump’s Action On The Issue
"I can’t picture any other president in American history who would have done what he did. He turned on a dime. He got it in his head that this was something that he wanted to do."
— HHS Secretary Robert F. Kennedy Jr.
🌿 Cannabis ETFs: 🌎 $YOLO 🇺🇸 $MSOS
marijuanamoment.net/there-wi…
AI can appear like a black box: energy in, intelligence out.
Efficiency spans the stack, from data-center power and heat to clusters and workloads.
In 2025, hardware measures alone helped Nebius avoid 100+ GWh, equal to a year’s electricity for 9,800+ US households.
Learn more. 26806167.fs1.hubspotusercont…
