#TNGTech
LightNews LightNews
tngtech.com
Recap #tngbtd: Our 18th conference with 1,400 guests on-site, additional remote participants, 40 talks, and 74 speakers was a huge success! Recordings of most talks will be available soon at www.youtube.com/@tngtech. Thanks to everyone involved in this event. 🚀
October 28, 2025 at 10:24 AM
huggingfacetrends.bsky.social
今日のHuggingFaceトレンド

tngtech/DeepSeek-TNG-R1T2-Chimera
本リポジトリは、DeepSeekの複数の既存モデルを「Assembly of Experts」手法で統合して開発された、新しい大規模言語モデル「DeepSeek-TNG R1T2 Chimera」（671B）を公開することを目的としています。
先行モデルで課題となっていたトークンの一貫性問題を解決し、新たな知能と出力トークン長の最適なバランスを提示しています。
モデルの仕様、構築方法、および詳細な評価結果が共有されています。
tngtech/DeepSeek-TNG-R1T2-Chimera · Hugging Face
We’re on a journey to advance and democratize artificial intelligence through open source and open science.
huggingface.co
July 10, 2025 at 10:19 AM
winbuzzer.com
New DeepSeek-R1T-Chimera Model Merges R1 Reasoning With Efficiency of V3-0324

#AI #LLMs #DeepSeekR1 #DeepSeekV3 #Chimera #OpenSourceAI #TNGTech #MoE #MachineLearning #TechNews #GenAI

winbuzzer.com/2025/04/27/n...
New DeepSeek-R1T-Chimera Model Merges R1 Reasoning With Efficiency of V3-0324 - WinBuzzer
DeepSeek-R1T-Chimera is a 685B MoE model built from DeepSeek R1 and V3-0324, focusing both on reasoning and performance.
winbuzzer.com
April 27, 2025 at 1:02 PM
jandotai.bsky.social
DeepSeek-R1T-Chimera hits R1-level reasoning, runs much faster, and uses ~40% fewer output tokens. It's built by merging routed experts from R1 and V3. huggingface.co/tngtech/Dee...
tngtech/DeepSeek-R1T-Chimera · Hugging Face
huggingface.co
April 28, 2025 at 1:28 PM
deepseek.activitypub.awakari.com.ap.brid.gy
DeepSeek-TNG-R1T2-Chimera Article URL: https://huggingface.co/tngtech/DeepSeek-TNG-R1T2-Chimera Comments URL: https://news.ycombinator.com/item?id=44449540 Points: 3 # Comments: 0

Origin | Interest | Match
tngtech/DeepSeek-TNG-R1T2-Chimera · Hugging Face
We’re on a journey to advance and democratize artificial intelligence through open source and open science.
huggingface.co
July 2, 2025 at 10:59 PM
tngtech.com
openrouter.ai/tngtech/deep...
DeepSeek R1T Chimera (free) - API, Providers, Stats
DeepSeek-R1T-Chimera is created by merging DeepSeek-R1 and DeepSeek-V3 (0324), combining the reasoning capabilities of R1 with the token efficiency improvements of V3. It is based on a DeepSeek-MoE Tr...
openrouter.ai
May 2, 2025 at 6:15 AM
tngtech.com
Serving #LLMs for over 50 applications, thereby consuming more than 100M tokens while generating over 10M tokens/day, requires us to carefully tune our request processing. Have a look at our article on @hf.co 👉 huggingface.co/blog/tngtech...
June 12, 2025 at 8:56 AM
huggingfacetrends.bsky.social
今日のHuggingFaceトレンド

tngtech/DeepSeek-TNG-R1T2-Chimera
このリポジトリは、複数の既存のDeepSeekモデルを統合・改良した新しいテキスト生成向け大規模言語モデル「DeepSeek-TNG-R1T2-Chimera」を公開するものです。
以前のモデルの課題を解決し、知能と出力効率を向上させることを目的としています。
tngtech/DeepSeek-TNG-R1T2-Chimera · Hugging Face
We’re on a journey to advance and democratize artificial intelligence through open source and open science.
huggingface.co
July 8, 2025 at 10:19 AM
sungkim.bsky.social
There are a few proposed solutions:

- TNG Tech's DeepSeek-R1T-Chimera ( huggingface.co/tngtech/Deep... )
- Moonshot AI's long2short methods as documented in Kimi k1.5: Scaling Reinforcement Learning with LLMs ( arxiv.org/abs/2501.12599 )
April 29, 2025 at 6:14 PM
tngtech.com
Model weights: huggingface.co/tngtech/olmO...
tngtech/olmOCR-7B-faithful · Hugging Face
We’re on a journey to advance and democratize artificial intelligence through open source and open science.
huggingface.co
April 23, 2025 at 10:16 AM
huggingfacetrends.bsky.social
今日のHuggingFaceトレンド

tngtech/DeepSeek-TNG-R1T2-Chimera
このリポジトリは、DeepSeekの複数の親モデル（R1-0528、R1、V3-0324）を「Assembly of Experts」手法で統合して構築された、新しい大規模言語モデル「DeepSeek-TNG R1T2 Chimera」（671B）を公開することを目的としています。
このモデルは、以前のモデルの課題を解決し、知能と出力トークン長のバランスを改善した最先端の言語生成能力を提供します。
tngtech/DeepSeek-TNG-R1T2-Chimera · Hugging Face
We’re on a journey to advance and democratize artificial intelligence through open source and open science.
huggingface.co
July 6, 2025 at 10:18 AM
nco.dev
if you want to be excited about something, check out

huggingface.co/tngtech/Deep...

they merged Deepseek R1 and v3 and found that it got much smarter than v3 alone, but without COT - and according to some people that have tried it, it's vibes check out

It might be on some API providers soon
tngtech/DeepSeek-R1T-Chimera · Hugging Face
We’re on a journey to advance and democratize artificial intelligence through open source and open science.
huggingface.co
April 28, 2025 at 6:13 AM
tngtech.com
Serving #LLMs to multiple applications and users is challenging due to limited GPU resources. Our colleague Benjamin Merkel explores common issues in request queueing and potential solutions such as fair scheduling based on metrics in his article: huggingface.co/blog/tngtech...
April 4, 2025 at 10:45 AM
bluesky.awakari.com
DeepSeek-TNG-R1T2-Chimera Article URL: https://huggingface.co/tngtech/DeepSeek-TNG-R1T2-Chimera Comments URL: https://news.ycombinator.com/item?id=44449540 Points: 3 # Comments: 0

Interest | Match | Feed
Origin
huggingface.co
July 2, 2025 at 10:58 PM
doagev.bsky.social
Thomas Endres und das @tngtech Team nehmen euch bei der #KINAVIGATOR mit auf eine Reise durch die Welt der KI: Neuronale Netze zur Erzeugung von Kunst, Deepfakes in Echtzeit, ein NLP-Chatbot und eine KI, die Social-Media-Kommentare generiert. https://scomp.ly/dBaxle8
November 27, 2024 at 7:22 PM
timkellogg.me
R1 Chimera: a model merge of the routed experts of DeepSeek R1 and V3

The resulting merged model performs as well as R1 but without the wandering thought traces. Just as smart, but faster.

huggingface.co/tngtech/Deep...
A scatter plot titled “Intelligence Score vs. Inference Cost” is shown, with the TNG Technology Consulting logo in the top left. The x-axis represents “Inference cost in % of R1 output tokens” and ranges from 30 to 100. The y-axis represents “Intelligence score (AIME 24 & MT-Bench)” and ranges from 70 to 90. Four points are plotted: • V3 in green near (40, 75) • R1T in light blue near (70, 87) • R1 in red at (100, 85) • A light blue label “R1T Chimera [V3+R1]” next to R1T • DeepSeek V3-0324 in green, associated with V3 • DeepSeek R1 in red, associated with R1 Two blue arrows labeled “smarter” and “faster” point respectively upward and leftward toward the R1T point, suggesting that R1T is both smarter (higher intelligence score) and faster (lower inference cost) compared to R1. At the bottom left, the Twitter handle @tngtech is present. The overall design is clean, with the plot emphasizing how models compare in terms of intelligence and efficiency.
April 27, 2025 at 11:29 AM
hikingdude.mastodon.social.ap.brid.gy
The @tngtech #bigtechday was amazing. Met a couple of old and new faces.

The talk of the #schwarzdigits group was pretty cool. They sound pretty mature.

Yet I noticed that I'm not so good in getting in touch with new people on such a big event. 😕
October 24, 2025 at 6:32 PM
tngtech.com
At TNG, we handle 5,000+ #LLM requests per hour and generate 10+ million tokens every day. Learn how our team optimizes inference serving for low-latency responses in high-traffic environments in the second article of our series on LLM performance: huggingface.co/blog/tngtech...
April 16, 2025 at 10:19 AM
aime-hq.bsky.social
TNG Tech created a LLM-Chimera by merging DeepSeek-R1 and DeepSeek-V3, combining the reasoning capabilities of R1 with the token efficiency improvements of V3.

In benchmarks, it appears to be as smart as R1 but much faster, using 40% fewer output tokens.

huggingface.co/tngtech/Deep...
tngtech/DeepSeek-R1T-Chimera · Hugging Face
We’re on a journey to advance and democratize artificial intelligence through open source and open science.
huggingface.co
April 29, 2025 at 2:56 PM
moonlitism.lesbian.cat
siga este tutorial

www.reddit.com/u/imowlekk/s...

são 200 msg grátis por dia

os modelos do deepseek sao

deepseek-ai/DeepSeek-R1

deepseek-ai/DeepSeek-R1-0528

tngtech/DeepSeek-R1T-Chimera

microsoft/MAI-DS-R1-FP8

deepseek-ai/DeepSeek-V3-0324
deepseek-ai/DeepSeek-V3
deepseek-ai/DeepSeek-V3-Base
no strawpage: como ce fez pra instalar o proxy do deepseek no janitor?
July 1, 2025 at 6:27 PM
hnws.bsky.social
DeepSeek-TNG-R1T2-Chimera
L: https://huggingface.co/tngtech/DeepSeek-TNG-R1T2-Chimera
C: https://news.ycombinator.com/item?id=44449540
posted on 2025.07.02 at 18:32:17 (c=0, p=4)
July 2, 2025 at 11:31 PM
tngtech.com
We recently fine-tuned an Optical Character Recognition AI model based on #olmOCR to help us automate our document processing workflows. In an article on Hugging Face, we discuss how we trained the Vision Language Model and also share the model weights. #OCR #VLM

huggingface.co/blog/tngtech...
Finetuning olmOCR to be a faithful OCR-Engine
A Blog post by TNG Technology Consulting GmbH on Hugging Face
huggingface.co
April 23, 2025 at 10:16 AM
tngtech.com
huggingface.co/tngtech/Deep...
tngtech/DeepSeek-R1T-Chimera · Hugging Face
We’re on a journey to advance and democratize artificial intelligence through open source and open science.
huggingface.co
May 2, 2025 at 6:15 AM
sungkim.bsky.social
It is not a finetune or distillation, but constructed from neural network parts of both parent MoE models.

huggingface.co/tngtech/Deep...
tngtech/DeepSeek-R1T-Chimera · Hugging Face
We’re on a journey to advance and democratize artificial intelligence through open source and open science.
huggingface.co
April 27, 2025 at 7:47 PM