Skip to main content
Live
Main content

Category

Models

Foundation models, benchmarks, and the research race.

More in Models

Nvidia logo
Models

Former OpenAI researcher finds Blackwell fingerprints in six Chinese AI models

Will DePue's forensic report says DeepSeek, Moonshot, MiniMax and Ant Group models released since April 2026 were initialized on export-controlled Nvidia GPUs.

Jaeden Schafer5 min read
Google logo
Models

Google launches Gemini Decisions API for fast closed-set AI decisions

The new Gemini Decision 1 model answers up to 1,000 yes-or-no, choice or rubric questions per request, billed at $0.10 per million input tokens.

Jaeden Schafer4 min read
Google logo
Models

Google DeepMind uses AlphaProof Code to verify Signal protocol at machine-code level

Gemini-powered agents produced about 6.5 million lines of machine-checked Lean proof in four weeks, and the Lean artifacts are being open-sourced.

Jaeden Schafer5 min read
Google logo
Models

Google opens Gemini 4 Carbon to developers at $2 per million input tokens

Google's strongest model moves past its cyber-defender preview to paying API customers, enterprises and AI Ultra subscribers, with prices doubling after Nov. 30.

Jaeden Schafer5 min read
Meta logo
Models

Meta launches Muse Ember, a cheap small model for agent work

Priced at $0.10 per million input tokens, Ember ties Claude Haiku 5.5 on the Artificial Analysis index and ships with a 1M-token context window.

Jaeden Schafer5 min read
Anthropic logo
Models

Anthropic launches Claude Fable 5.2 and Mythos 5.2, cuts Fable price 20%

Fable 5.2 ships to all users with outputs up to 30% faster, while Mythos 5.2 stays limited to vetted cyber and life sciences organizations.

Jaeden Schafer5 min read
Google logo
Models

Google unveils TPU 8tx and TPU 8i Max for confidential and long-context AI

The two new eighth-generation systems make four TPU launches in 2026, a first for Google, as TPU customers near 100.

Jaeden Schafer5 min read
Anthropic logo
Models

Anthropic details how Claude agents dream to clean up their memory

In Anthropic's CL-Bench test, a memory store lifted Claude Opus 5.5 from 0.655 to 0.838, and idle-time review helped fix bad notes.

Jaeden Schafer5 min read
Anthropic logo
Models

PSI says a non-public Anthropic model proved the 2008 Aaronson-Ambainis conjecture

The Physical Superintelligence system had a complete, Lean-checked proof on day three of its run, and Stanford researchers are now reviewing it.

Jaeden Schafer4 min read
Anthropic logo
Models

METR finds Anthropic's model progress has roughly doubled, mostly due to AI

An 11-week high-access evaluation puts the speedup at 1.9x since October 2025 and attributes perhaps 80% of it to Anthropic's own use of AI.

Jaeden Schafer5 min read
Thinking Machines debuts Inkling 2 Small, a 594B MoE model, on Tinker
Models

Thinking Machines debuts Inkling 2 Small, a 594B MoE model, on Tinker

The model runs 17B active parameters, and the lab says Apache 2.0 weights follow next week once external red-teaming wraps.

Jaeden Schafer5 min read
Baidu releases ERNIE 5.5 Flash Base, a 280B open-weight MoE model
Models

Baidu releases ERNIE 5.5 Flash Base, a 280B open-weight MoE model

The pre-training checkpoint activates 14B parameters per token, was trained on 33.5T tokens and ships with a 1M-token context configuration.

Jaeden Schafer4 min read
Google logo
Models

Google ships Gemini Omni 1.2 Flash with 15-second video and a 30% price cut

The video model adds multi-turn conversational editing and tighter reference following, priced at $0.07 per second of 720p video across the Gemini API and Cloud.

Jaeden Schafer5 min read
OpenAI logo
Models

OpenAI launches ChatGPT Images 2.6 with animated GIFs and faster API models

Two new developer models, Flare and Sunburst, arrive alongside reusable reference images, image markup, SVG export and up to 50% faster generation.

Jaeden Schafer3 min read
Cognition launches SWE-2 Flash at one-fifth the price of SWE-2
Models

Cognition launches SWE-2 Flash at one-fifth the price of SWE-2

The new Devin coding model runs three times faster than SWE-2, is built on Thinking Machines' Inkling-Small, and is free through November 30.

Jaeden Schafer4 min read
Tencent releases Hy4, filtering exploit and pen-testing data out of training
Models

Tencent releases Hy4, filtering exploit and pen-testing data out of training

The official Hy4 adds vision and computer use, cuts its internal hallucination rate to 7.2%, and keeps the preview's API price.

Jaeden Schafer5 min read
OpenAI logo
Models

OpenAI says GPT-6 Astra designed a Codex hardware prototype in under a week

Astra handled the enclosure, circuit board, firmware and interface; a contract manufacturer delivered prototypes in under a month after several revision rounds.

Jaeden Schafer4 min read
Meituan releases LongCat-2.5, a 1.6T-parameter multimodal model for agents
Models

Meituan releases LongCat-2.5, a 1.6T-parameter multimodal model for agents

The model activates about 48B parameters per token and scored 91.9 on BrowseComp, according to results Meituan reported.

Jaeden Schafer4 min read
SemiAnalysis launches GDPx, a benchmark where the best AI model scores 12%
Models

SemiAnalysis launches GDPx, a benchmark where the best AI model scores 12%

The 11-domain agentic test runs models for up to 8 hours in simulated work environments; Claude Opus 5.5 is the best all-around performer.

Jaeden Schafer5 min read
OpenAI logo
Models

OpenAI releases GPT-Rosalind-6, a biology model on GPT-6 Astra weights

Same weights as GPT-6 Astra, fewer dual-use refusals and a jump to 49.6% on MedChemBench, but only for vetted labs through trusted access.

Jaeden Schafer4 min read
OpenAI logo
Models

OpenAI rewrites its Model Spec around five values and renames the assistant GPT

Most rules become defaults that developers and users can change, detailed style guidance is cut, and hard limits stay only where OpenAI deems them necessary.

Jaeden Schafer5 min read
Mistral releases open weights for Large 4, its 1-trillion-parameter model
Models

Mistral releases open weights for Large 4, its 1-trillion-parameter model

The checkpoint lands on Hugging Face three weeks after preview, scoring 41% on Terminal-Bench 4.0 and priced at $0.70 per million input tokens via API.

Jaeden Schafer5 min read
OpenAI logo
Models

DeepSeek and Moonshot trained new models on OpenAI's GPT-6 Astra, report says

The report names Kimi K3.1 Pro and DeepSeek V4.1-Flash 1017, a month after OpenAI tied a reasoning-extraction campaign to Moonshot.

Jaeden Schafer4 min read
Google logo
Models

Google unveils Gemini 4 Carbon, ranking first on Artificial Analysis index at 59

Carbon arrives two weeks after Gemini 4 Argon, goes first to cyber defenders, and keeps Argon's $2/$10 introductory per-million-token pricing.

Jaeden Schafer5 min read
AI Box Daily briefingFree · Daily · No fluff

Stay ahead of everyone in AI.

The tightly edited AI news email engineers, founders, and investors actually open. One email. Every weekday. Five minutes to finish.

Loved by 10,000+ AI professionals
Free forever. Unsubscribe with one click.

The briefing read inside teams at