ai-models - The LGTM (Page 6)

The LGTM

Home
Agentic Coding
Claude Code
Codex

Sign in Subscribe

ai-models

A collection of 70 posts

Xiaomi Launches Three MiMo-V2 AI Models — Trillion-Parameter Flagship, Multimodal, and TTS

Xiaomi Launches Three MiMo-V2 AI Models — Trillion-Parameter Flagship, Multimodal, and TTS

Xiaomi officially launched the MiMo-V2 family — three models in one shot: MiMo-V2-Pro (1 trillion total parameters, 42B active via MoE, 1M token context), MiMo-V2-Omni (multimodal agent model), and MiMo-TTS (expressive speech synthesis). MiMo-V2-Pro ranks 7th globally on the Artificial Analysis Intelligence Index and 3rd on both PinchBench and ClawEval — trailing

23 Mar 2026 1 min read

Cursor Admits Its New Composer 2 Coding Model Was Built on Moonshot AI's Kimi K2.5

Cursor's newly launched Composer 2 — marketed as the company's most capable coding model yet — turns out to be built on top of Kimi K2.5, an open-source model from Chinese AI lab Moonshot AI. Community developers spotted the underlying model ID before Cursor officially confirmed it.

23 Mar 2026 1 min read

Mistral Launches Forge — Enterprise Model Training Platform — Plus Mistral Small 4 and NVIDIA Coalition

Mistral AI packed a remarkable amount into a single week. The Paris-based lab launched Forge, a new enterprise platform designed to take organizations through the full lifecycle of building and owning proprietary AI models — from pre-training on their own data all the way through reinforcement learning alignment. It's

22 Mar 2026 1 min read

Anthropic Drops Long-Context Pricing Surcharge — 1M Tokens Now at Standard Rate

Anthropic Drops Long-Context Pricing Surcharge — 1M Tokens Now at Standard Rate

Anthropic quietly removed one of the more significant cost barriers in enterprise AI last week: the long-context pricing surcharge on Claude Opus 4.6 and Sonnet 4.6 is gone. As of March 14th, processing up to one million tokens now costs exactly the same per-token rate as any standard

22 Mar 2026 1 min read

Anthropic vs. Pentagon: Court Filings Reveal DoD Privately Said Sides Were "Nearly Aligned"

Anthropic vs. Pentagon: Court Filings Reveal DoD Privately Said Sides Were "Nearly Aligned"

A surprising piece of the Anthropic-Pentagon dispute came into sharper focus this week as sworn court declarations painted a very different picture than the one the DoD presented publicly. Anthropic's Head of Policy Sarah Heck and Head of Public Sector Thiyagu Ramasamy both filed statements attesting that a

22 Mar 2026 1 min read

DeepSeek V4 Leaks Point to April Launch -- 1T Parameters, 40% Memory Reduction

DeepSeek V4 Leaks Point to April Launch -- 1T Parameters, 40% Memory Reduction

TL;DR: Spec leaks circulating in Chinese media detail DeepSeek V4 as a 1-trillion-parameter model with a 1M token context window, expected in April 2026. Engineers reportedly achieved a 40% memory reduction vs V3 and 1.8x inference speedup -- meaning organizations could run a trillion-parameter model on significantly less

22 Mar 2026 1 min read

The Quiet 48 Hours That Redrew the AI Map: March 19-21 Roundup

The Quiet 48 Hours That Redrew the AI Map: March 19-21 Roundup

TL;DR: Despite no flagship launches from OpenAI, Google, Anthropic, or Meta in the March 19-21 window, the AI community was buzzing. The Xiaomi/Hunter Alpha reveal, GPT-5.4 mini going free, and a wave of agent-focused open source tools dominated discourse. The piece argues the real story is a

22 Mar 2026 1 min read

White House Proposes National AI Framework -- States Would Lose Power to Regulate

White House Proposes National AI Framework -- States Would Lose Power to Regulate

TL;DR: The White House released a four-page national AI policy framework calling for broad federal preemption of state AI laws, no new federal oversight body, and no 'open-ended liability' for AI companies. The plan also requests streamlined permitting for data centers and regulatory sandboxes. House Republican leadership

22 Mar 2026 1 min read

OpenAI Gives Everyone Free Access to GPT-5.4 Mini Reasoning

OpenAI Gives Everyone Free Access to GPT-5.4 Mini Reasoning

TL;DR: OpenAI launched GPT-5.4 mini and nano, making the mini variant available to all free ChatGPT users via the 'Thinking' toggle. GPT-5.4 mini delivers improved coding and reasoning capabilities that closely approach the flagship GPT-5.4 model. For paid users, it serves as the automatic

22 Mar 2026 1 min read

Xiaomi's Stealth 1-Trillion-Parameter Hunter Alpha Revealed as MiMo-V2-Pro

TL;DR: Xiaomi's MiMo-V2-Pro, a 1-trillion-parameter model that appeared anonymously on OpenRouter as 'Hunter Alpha,' has been officially revealed. Built by former DeepSeek researcher Luo Fuli, the model surpassed 1 trillion tokens in usage during its stealth phase, topped OpenRouter's usage charts, and benchmarks

22 Mar 2026

← Newer Posts Page 6 of 6

The LGTM © 2026

Sign up

Powered by Ghost