corpus.blog
41554 blogs · 4,417,066 posts Claim your blog
Record · corpus.blog/posts/01a07df3-ecc9-70ed-9df1-2ee53539c8a0

From DeepSeek V3 to V3.2: Architecture, Sparse Attention, and RL Updates

Ahead of AI · published 3 December 2025

A record is what was published and who pointed at it. The text of the post is not held here: read it at the source, or at its Wayback capture.

Read the post at magazine.sebastianraschka.com Wayback capture
2blogs citing
First cited
31 Dec 2025
Most recent
18 Jun 2026
Rank this month
not ranked
Links from this post
44
Words captured
6,031

Cited by

Blog
In the post
Date
Sebastian Raschka
GLM-5.2 and IndexShare for Long-Context Sparse Attention Read at the source ↗
18 Jun 2026
smw.ai
State of LLMs 2025 Read at the source ↗
31 Dec 2025

Links from this post

Link
Host
https://substackcdn.com/image/fetch/$s_!JN06!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Ff841a4f1-8524-4e3a-98f2-b9bf65072819_2113x1459.png
substackcdn.com
https://huggingface.co/deepseek-ai/DeepSeek-V3.2/blob/main/assets/paper.pdf
huggingface.co
The Big LLM Architecture Comparison Read at the source ↗
magazine.sebastianraschka.com
https://substackcdn.com/image/fetch/$s_!W4Qo!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Ff4b97110-d705-4531-b4af-4f87187a8dea_1393x1394.png
substackcdn.com
https://substackcdn.com/image/fetch/$s_!4LZs!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F56649604-b822-47a2-a3fc-169ec453ae17_1414x446.png
substackcdn.com
https://x.com/rasbt/status/1973043285624291787?s=20
x.com
https://substackcdn.com/image/fetch/$s_!maHo!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F269735af-5f6c-4a6f-b9f4-56b6d2d4d41f_1754x1322.png
substackcdn.com
Understanding Reasoning LLMs Read at the source ↗
magazine.sebastianraschka.com
https://substackcdn.com/image/fetch/$s_!qqkY!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fda512f7b-5a41-49c1-8a76-ff74fa065972_1145x928.png
substackcdn.com
https://magazine.sebastianraschka.com/i/168650848/mixture-of-experts-moe
magazine.sebastianraschka.com
https://arxiv.org/abs/2412.19437
arxiv.org
https://substackcdn.com/image/fetch/$s_!8R3i!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F5d55cf9d-7855-4802-b7de-61971cd02f38_1623x1527.png
substackcdn.com
https://arxiv.org/abs/2405.04434
arxiv.org
https://substackcdn.com/image/fetch/$s_!12jb!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F4848890b-e00b-468f-880d-27ee5db5fb3a_1277x619.png
substackcdn.com
https://substackcdn.com/image/fetch/$s_!dNxH!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fe798b5b2-b489-485f-8b89-6043f3029d02_2019x1142.png
substackcdn.com
The State of Reinforcement Learning for LLM Reasoning Read at the source ↗
magazine.sebastianraschka.com
https://huggingface.co/deepseek-ai/DeepSeek-R1-0528
huggingface.co
https://github.com/deepseek-ai/DeepSeek-V3.2-Exp/blob/main/DeepSeek_V3_2.pdf
github.com
https://substackcdn.com/image/fetch/$s_!IwOF!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F0a4f4aa7-39fb-4c7d-8c25-a7643475a1c6_799x418.png
substackcdn.com
https://substackcdn.com/image/fetch/$s_!oxpT!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Faabfcdab-fd13-44f8-94d9-e8fd27cd48b4_815x424.png
substackcdn.com
https://huggingface.co/deepseek-ai/DeepSeek-V3.2-Exp/blob/main/inference/model.py
huggingface.co
https://substackcdn.com/image/fetch/$s_!tSrP!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F763113a8-3405-4c48-9d24-35a35bf29e4e_1388x928.png
substackcdn.com
https://arxiv.org/abs/2511.22570v1
arxiv.org
https://arxiv.org/abs/2211.14275
arxiv.org
https://arxiv.org/abs/2305.20050
arxiv.org
https://substackcdn.com/image/fetch/$s_!MVTj!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fef977cde-cec9-4f7c-a805-b450d01f9e54_1907x1195.png
substackcdn.com
https://substackcdn.com/image/fetch/$s_!ITiX!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fe8da9860-f197-4f9f-8af9-aa055de71dc5_2404x1397.png
substackcdn.com
https://substackcdn.com/image/fetch/$s_!HFcU!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F5899b054-e704-47a0-85e4-58dc217db84a_1907x1494.png
substackcdn.com
https://substackcdn.com/image/fetch/$s_!shTj!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F5a02acb6-813c-40f0-89ed-0b1025a3c465_1907x1494.png
substackcdn.com
https://substackcdn.com/image/fetch/$s_!Cmee!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F89a882b5-ebca-442f-9bef-1a5ef1b4c570_1619x992.png
substackcdn.com
https://mng.bz/Nwr7
mng.bz
https://huggingface.co/deepseek-ai/DeepSeek-V3.2/resolve/main/assets/paper.pdf
huggingface.co
https://substackcdn.com/image/fetch/$s_!dm8C!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F08cd373b-6a67-429e-82ba-dc5758841b38_809x1336.png
substackcdn.com
https://substackcdn.com/image/fetch/$s_!j_TQ!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F67afa7ab-07b8-4e78-afb6-136ae0b6fa54_1445x804.png
substackcdn.com
https://arxiv.org/abs/2503.14476
arxiv.org
https://arxiv.org/abs/2503.20783
arxiv.org
https://www.datocms-assets.com/64837/1763662397-1763646865-olmo_3_technical_report-1.pdf
datocms-assets.com
https://substackcdn.com/image/fetch/$s_!QaOn!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Ff87eee11-3aaa-41c4-b45a-51b5c865a5ae_1131x574.png
substackcdn.com
https://arxiv.org/abs/2512.24880
arxiv.org
https://arxiv.org/abs/2409.19606
arxiv.org
https://substackcdn.com/image/fetch/$s_!cQTW!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F7e9fe8ab-954c-4495-9150-0fb67176c5bd_1890x1352.png
substackcdn.com
https://amzn.to/4fqvn0D
amzn.to
https://substackcdn.com/image/fetch/$s_!RCl_!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F27a118a0-5da6-4486-b1f0-3743754d0a77_8106x4044.webp
substackcdn.com
https://www.amazon.com/Build-Large-Language-Model-Scratch/dp/1633437167
amazon.com