/
Navigation
C
Chronicles
Browse all articles
C
E
Explore
Semantic exploration
E
R
Research
Entity momentum
R
N
Nexus
Correlations & relationships
N
~
Story Arc
Topic evolution
S
Drift Map
Semantic trajectory animation
D
P
Posts
Analysis & commentary
P
Browse
@
Entities
Companies, people, products, technologies
Domains
Browse by publication source
Handles
Browse by social media handle
Detection
?
Concept Search
Semantic similarity search
!
High Impact Stories
Top coverage by position
+
Sentiment Analysis
Positive/negative coverage
*
Anomaly Detection
Unusual coverage patterns
Analysis
vs
Rivalry Report
Compare two entities head-to-head
/\
Semantic Pivots
Narrative discontinuities
!!
Crisis Response
Event recovery patterns
Connected
Nav: C E R N
Search: /
Command: ⌘K
Embeddings: large
VOICE ARCHIVE

@gm8xx8

@gm8xx8
10 posts
2026-01-27
KIMI K2.5 VISUAL AGENTIC INTELLIGENCE AT 1T SCALE Kimi K2.5 is Moonshot's open-source successor to K2: ~15T mixed vision+text continual pretraining plus a real scale-out agent stack (Agent Swarm) across Instant / Thinking / Agent modes. KIMI K2.5: MoE backbone (1T total / 32B [image]
2026-01-27 View on X
Kimi

Moonshot says Kimi K2.5 builds on K2 with “pretraining over ~15T mixed visual and text tokens” and “can self-direct an agent swarm with up to 100 sub-agents”

Today, we are introducing Kimi K2.5, the most powerful open-source model to date.

KIMI K2.5 VISUAL AGENTIC INTELLIGENCE AT 1T SCALE Kimi K2.5 is Moonshot's open-source successor to K2: ~15T mixed vision+text continual pretraining plus a real scale-out agent stack (Agent Swarm) across Instant / Thinking / Agent modes. KIMI K2.5: MoE backbone (1T total / 32B [image]
2026-01-27 View on X
Bloomberg

Chinese startup Moonshot releases Kimi K2.5, saying the model can process text, images, and videos simultaneously and beats its open-source peers in some tests

Alibaba Group Holding Ltd.-backed Moonshot AI released an upgrade of its flagship model, heating up a domestic arms race ahead …

2026-01-15
Thinking Machines is moving with clarity and speed. I'm more bullish on them by the day. @soumithchintala is a rare kind of hire. The kind that quietly raises the ceiling of what a team can execute. Another strong move, IMO.
2026-01-15 View on X
Wired

A source close to Thinking Machines Lab alleges ex-CTO Barret Zoph, who is returning to OpenAI, had shared confidential company information with competitors

The departures are a blow for Thinking Machines Lab.  Two narratives are already emerging about why they happened.

Thinking Machines is moving with clarity and speed. I'm more bullish on them by the day. @soumithchintala is a rare kind of hire. The kind that quietly raises the ceiling of what a team can execute. Another strong move, IMO.
2026-01-15 View on X
@kyliebytes

Thinking Machines Lab parts ways with CTO Barret Zoph, with Soumith Chintala taking over the role; sources say the termination is due to “unethical conduct”

BREAKING: Thinking Machines has terminated its CTO, Barret Zoph, due to unethical conduct according to two sources familiar with the matter. CEO Mira Murati announced the news at a...

2025-12-07
RNJ-1 (8B BASE / INSTRUCT) Open-weight 8B dense Transformer trained from scratch by Essential AI, with fully specified architecture and training dynamics. Global self-attention with YaRN extending context to 32k tokens. Pre-training-first, compression-driven approach targeting [image]
2025-12-07 View on X
Essential AI

Essential AI, whose CEO co-wrote Google's Attention Is All You Need paper, unveils Rnj-1, an 8B-parameter open model with SWE-bench performance close to GPT-4o

The long-term advancement and equitable diffusion of AI technologies crucially depend on their development in the Open.

2025-10-01
GLM-4.6 is here. Zhipu extends the 4.5 line with: - Context: 128K → 200K tokens for more complex agent workflows - Coding: stronger results on benchmarks + real-world apps (Claude Code, Cline, Roo, Kilo), incl. polished front-end generation - Reasoning: better inference-time [image]
2025-10-01 View on X
Z.ai

Z.ai releases GLM-4.6, an open-weights model with a context window of up to 200K tokens, claiming near parity with Claude Sonnet 4 on coding and reasoning tasks

Today, we are releasing the latest version of our flagship model: GLM-4.6.  Compared with GLM-4.5, this generation brings several key improvements:

2025-04-30
Xiaomi MiMo-7B: a 7B reasoning model series trained from scratch, outperforms 32B+ baselines on math and code via dense RL - pretrained on 25T tokens w/ multi-token prediction - RL rewards from rule-verifiable math/code tasks - cold-start RL model (MiMo-7B-RL-Zero) hits 93.6% [image]
2025-04-30 View on X
Bloomberg

Xiaomi unveils open-source AI reasoning model MiMo, joining other Chinese tech leaders hoping to make a splash in the burgeoning AI field endorsed by Beijing

Xiaomi debuted MiMo a day after Alibaba unveiled the latest version of its own flagship model, amplifying a race between China's tech players …

2025-03-12
Gemma 3 is a family of open models (1B, 4B, 12B, 27B) designed for efficient, on-device use. They support 140 languages, text and visual reasoning, 128k-token context, function calling, and structured outputs. Quantized versions reduce compute and memory with minimal performance [image]
2025-03-12 View on X
9to5Google

Google unveils Gemma 3, the “world's best single-accelerator model”, running on a single GPU, in 1B, 4B, 12B, and 27B sizes, and says it outperforms Llama-405B

Following version 1 in February 2024 and 2 in May, Google today announced Gemma 3 as its latest open model for developers.

2025-01-28
Qwen2.5-VL is a highly capable vision-language model designed for a wide range of tasks. It excels in visual understanding, device interaction, and long-form video analysis. Key Features: - Visual Understanding: Handles everything from simple objects to complex data [image]
2025-01-28 View on X
TechCrunch

Alibaba's Qwen team releases Qwen2.5-VL, a new series of AI models that can control PCs and phones, as well as perform a number of text and image analysis tasks

QWEN CHAT GITHUB HUGGING FACE MODELSCOPE DISCORD Anusuya Lahiri / Benzinga : Not Just DeepSeek - Alibaba Unveils AI Model To Rival OpenAI's Operator Markus Kasanmascheff / WinBuzze...

2024-12-20
Gemini 2.0 Flash Thinking Experimental > Multimodal Understanding: Handles tasks involving multiple data types. > Reasoning Capabilities: Excels at solving complex problems. > Coding: Tackles difficult code and math challenges. > Visible Thinking Process: Shows the model's
2024-12-20 View on X
TechCrunch

Google releases Gemini 2.0 Flash Thinking, an experimental “reasoning” model that “explicitly shows its thoughts” and can use them to strengthen its reasoning

Quick: what sort of prompts should you run against GPT-4o vs Gemini 1.5 Flash vs o1 vs o1-pro vs gemini-2.0-flash-thinking-exp? X: Jeff Dean / @jeffdean : Introducing Gemini 2.0 Fl...