Models

69 articles

Models

Clawpit

Google tests direct Flipkart purchases inside Gemini and AI Mode in India

Google has started an experiment that lets users in India buy Flipkart products directly inside the Gemini and AI Mode interfaces — no redirect to a t…

ModelsAuto-updating

AI drives $942 million jump in health-care bills over two years
An analysis by the Blue Cross Blue Shield Association (BCBSA) found that hospitals’ use of artificial-intelligence tools to file insurance claims adde

Models27.09.2026New
Matthew Prince tries to save the web from AI bots
The bots are taking over traffic. Cloudflare data from June shows more than half of all web traffic now comes from bots, and the share keeps climbing

Models26.09.2026New
Meta tries to turn Muse into a tiny tax on global e-commerce
The economic logic behind Meta's Muse does not rest on $20 or $100 monthly subscriptions, nor on advertising. The direction is a transaction fee — a s

Models26.09.2026New
Qwen Image 3 Pro takes OpenRouter's first benchmark
OpenRouter just published results from its first public benchmark — a meme-generation contest where image models went head-to-head on a single prompt.

Models26.09.2026
Fastino Labs releases 340M-parameter decision model that runs on CPU and returns structured answers
with confidence scores Fastino Labs has released GLiNER2.5-Decide, a 340-million-parameter open-weight decision model that accepts text and a schema o

Models26.09.2026
OpenAI's Astra and Anthropic's Claude Opus 5 crack two long-unsolved Enigma messages
Two large language models — OpenAI's Astra (GPT-6) and Anthropic's Claude Opus 5 — have decrypted a pair of Nazi Enigma messages that had resisted res

Models25.09.2026
Meta's Muse exposes its full filesystem — on purpose
Muse, Meta's chatbot, began offering users a compressed download of its entire filesystem yesterday, root access included. Just a day earlier it had r

Models25.09.2026
Lightspeed halves India fund to $250M and goes all-in on AI
Lightspeed is targeting $250 million for its fifth India fund — half the size of the $500 million vehicle it closed in 2022 — and centering the entire

Models25.09.2026
BottleCap launches ThinkingCap-Qwen3.8-27B, a fine-tune that cuts reasoning tokens 37.2% on average
BottleCap has released ThinkingCap-Qwen3.8-27B, a fine-tune of Qwen3.8-27B whose sole purpose is to shorten the chain of thought. Across 12 benchmarks

Models25.09.2026
Google releases Gemini 3.8 Live with Live Avatar for enterprise customers
Google is rolling out Gemini 3.8 Live with live video avatar capability inside Gemini Enterprise, one week after announcing the model itself and its E

Models25.09.2026
Google lets Gemini call businesses for you with new Call for Me trial
Google has opened a limited trial that lets its Gemini model place real phone calls to businesses on a user's behalf. The feature, called Call for Me,

Models25.09.2026
Meta unveils Muse Charm AI pendant to take on the smart-pendant category
Mark Zuckerberg pulled a classic “one more thing” at the close of Meta’s annual Connect conference, revealing a pendant called Muse Charm. The device

Models24.09.2026
Meta’s Muse Charm looks like a Tamagotchi, but it’s riding a completely different trend
Meta has announced Muse Charm, a small pendant that at first glance recalls the digital pet of the nineties, but in practice it’s surfing a much more

Models24.09.2026
OpenRouter launches Memebench, a meme benchmark for language models
OpenRouter dropped Memebench yesterday. It's a new eval framework that tests whether large language models can riff on a meme template. Not factual re

Models24.09.2026
Shield AI, Waabi and GM to discuss building AI when failure isn't an option
A panel at TechCrunch Disrupt 2026 will tackle the gap between laboratory models and systems that must survive in the physical world. The session, "Bu

Models24.09.2026
New paper shows how to evaluate production agents at a fraction of the cost
Researchers reconstructed the full benchmark score for an active analytics agent serving tens of thousands of monthly users with just 38.5% of the que

Models24.09.2026
Gemini starts calling businesses for you on Pixel 11
Google is rolling out an early experiment that lets paid Gemini subscribers hand off phone calls to local businesses directly to the model. Instead of

Models24.09.2026
Nvidia's Jensen Huang defends data-center pollution with "painful surgery" argument
Nvidia chief executive Jensen Huang appeared on Ezra Klein's podcast and framed the interplay between artificial intelligence and the climate crisis i

Models24.09.2026
Liquid AI releases visual draft model for LFM2.5-VL-3B, claims up to 3.13× decoding speedup
Liquid AI has released LFM2.5-VL-DSpark, a draft model that attaches to LFM2.5-VL-3B to accelerate speculative decoding in vision-language models. The

Models24.09.2026
Nvidia and Google DeepMind release predicted protein structures for 2,800 viruses to open database
Nvidia, Google DeepMind, and the European Bioinformatics Institute (EMBL-EBI) have released predicted three-dimensional structures for protein complex

Models24.09.2026
AI learns to assemble furniture, changing the game
Researchers have launched the Furniture Assembly Benchmark (FAB), a test that asks vision models to spot errors in partially built IKEA furniture. The

Models24.09.2026
StrictlyVC returns to Disrupt with a focus on venture's new rulebook
The event runs at Moscone West in San Francisco from October 13 to 15, but the StrictlyVC track is concentrated entirely on the afternoon of the 14th,

Models24.09.2026
OpenAI launches open benchmark for mental health conversations, then tangles with its own categories
OpenAI released MentalHealthBench over the weekend, a first-of-its-kind open benchmark that attempts to cover the full spectrum of mental health conve

Models24.09.2026
Most workers use AI but get no time to learn it
A new Workera survey out this week shows 80% of large US companies believe they are on the right track for an AI-powered future, up from 67% a year ag

Models24.09.2026
Nvidia releases Nemotron 3 Diarization, a speaker-diarization model that tops the benchmark
Nvidia has released Nemotron 3 Diarization, a speaker-diarization model that identifies who spoke and when even when voices overlap, supports up to ei

Models24.09.2026
Viral tweet claimed Altman, Amodei led "Superintelligence" thanks — archive shows otherwise
A post by Nikolas Ioannou spread rapidly over the weekend asserting that the 2014 edition of Nick Bostrom's "Superintelligence" lists Sam Altman and D

Models24.09.2026
Google releases Gemini 3.8 Flash TTS and Flash-Lite TTS with prompt-based voice design
Google yesterday released two new text-to-speech models in the Gemini Audio family: Gemini 3.8 Flash TTS, built for deep creative direction and charac

Models24.09.2026
Anthropic hits 1.5x R&D acceleration, METR estimates
METR, an organization that specializes in evaluating model capabilities and risks, estimates that Anthropic has reached a 1.5x acceleration rate in AI

Models24.09.2026
Anthropic says its wet lab has already produced a significant discovery
Anthropic confirmed this week that the wet lab it set up in the San Francisco Bay Area last spring has identified an entirely new enzyme system the co

Models24.09.2026
Meta Connect 2026: VR glasses at 100 grams, a dedicated Muse device, and movies on Quest
The glasses that replace the headset Mark Zuckerberg closed the annual conference with an announcement that rewrites the VR form factor: new VR glasse

Models24.09.2026
Space Bunny Alpha returns to OpenRouter after outage
The anonymous model known as Space Bunny Alpha is back on OpenRouter after a temporary shutdown. It is a "stealth model" — developed and operated by a

Models23.09.2026
Vatican AI adviser warns major labs operate like a cartel
Paolo Benanti, a priest who advises the Catholic Church on artificial intelligence, argues that dramatic warnings about superintelligence are divertin

Models23.09.2026
GPT-6 automates 20.8% of remote-work projects, but that still sits near the floor
A jump from 2.5% to 20.8% automation of real projects sounds impressive, yet the new Remote Labor Index shows advanced models still fail the vast majo

Models22.09.2026
Grok 4.7 hits 60 percent on Vals Index but task cost nearly triples
The latest SDK update delivered a solid performance bump for Grok 4.7, lifting it from 54 percent to just over 60 percent on the Vals Index. The gain

Models22.09.2026
Tsinghua and Tencent researchers unveil a frozen vision-language model that drives robots
with simple motion commands RoboDawn hits 73.6% one-shot success on RoboTwin 2.0 C2R, beating policies trained on the full RoboTwin robot dataset — wi

Models22.09.2026
Claude Opus 5.5 tops Vals AI RSI benchmark, beats human baseline in LM Training
According to Vals AI, Claude Opus 5.5 has taken first place on the RSI benchmark and become the first model to surpass the published human reference o

Models22.09.2026
Artificial Analysis adds GPT-6 Sol and Luna to independent benchmark
The independent evaluation platform Artificial Analysis has updated its leaderboards to include GPT-6 Sol and Luna, the two newest models from OpenAI,

Models22.09.2026
Trump tells the UN: AI is now "super intelligence"
The announcement came Tuesday at the General Assembly. President Trump declared the United States is officially changing the term "artificial intellig

Models22.09.2026
Amazon opens access to video model developed from rat neural activity
Select Amazon Web Services customers will receive early access starting Tuesday to a video model from The Biological Computing Company, a startup buil

Models22.09.2026
Austrian Academy of Sciences releases Apollo, first advanced LLM built for Ancient Greek
The Austrian Academy of Sciences will release Apollo on Wednesday, the first advanced large language model developed specifically for Ancient Greek. B

Models22.09.2026
OpenAI forms mathematics advisory board after claiming Navier-Stokes solution
On Monday OpenAI announced the Advisory Group on Mathematics and Artificial Intelligence, an independent body that will sit at the Institute for Advan

Models22.09.2026
White Circle releases Halo 1.0, open-source training framework claiming 2.8x throughput on B300
White Circle has shipped version 1.0 of Halo, an open-source training framework that runs Hugging Face models without breaking their native format. Th

Models22.09.2026
Moonshot AI launches Kimi K3 on Amazon Bedrock with 1 million token context window
Moonshot AI's new open-weights model, Kimi K3, landed on Amazon Bedrock on September 18, days after its official unveiling. It brings a 1 million toke

Models22.09.2026
Egypt shifts from AI potential to production as Grand Egyptian Museum event signals step change
Nvidia leads the infrastructure from the top At the Grand Egyptian Museum yesterday, Egypt's AI builders — startups, researchers, developers and organ

Models22.09.2026
Alibaba launches Qwen-Image-2.1: unified 7B model for image generation and editing
A third of the size, twice the jobs Alibaba's Qwen team has released Qwen-Image-2.1, an open-weight model that merges text-to-image generation and edi

Models21.09.2026
The virtual wall stops no one. It only counts the dead.
An investigation by MIT Technology Review, in partnership with the Times of San Diego, reveals that the smart surveillance towers strung along the U.S

Models21.09.2026
AI safety center exposes systematic cheating across every frontier model
The Center for AI Safety has released CheatBench, a benchmark that measures not how smart a model is but how willing it is to cheat to complete a task

Models21.09.2026
Nvidia highlights three companies using AI to accelerate clean energy at New York Climate Week
Nvidia used New York Climate Week to showcase five companies blending AI into clean-energy infrastructure, though the detailed briefing named only thr

Models21.09.2026
4B model hits 61.5% on SWE-bench Verified without distillation
A 4-billion-parameter model has hit 61.5% on SWE-bench Verified without distilling from a larger frontier model. Only a year ago, this level of perfor

Models21.09.2026
StepFun unveils Step 5 Preview, a 600B-parameter MoE model with a 1M-token context window
StepFun has launched Step 5 Preview, a flagship model built for long-horizon agentic work, software engineering, professional knowledge tasks and fina

Models21.09.2026
AI lab leaders talk about slowing down, but the details are missing
A public debate flared this week over whether the AI industry is actually prepared to tap the brakes. Dario Amodei, Anthropic's chief executive, publi

Models21.09.2026
Qwen releases Qwen-Image-2.1, a 7B open-weight model unifying generation and editing
Alibaba's Qwen team has released Qwen-Image-2.1, the latest in its image model series. The model uses a 7-billion-parameter architecture — compact by

Models20.09.2026
Open models reach biology but serving them is another story
Phylo, the applied research lab spun out of Stanford's Biomni open-source project, is moving its open-model serving layer to Fireworks AI. The shift c

Models19.09.2026
Linkup releases SPARSEUP, a 149M-parameter sparse encoder under Apache 2.0
Linkup's research team has open-sourced SPARSEUP, a sparse embedding model built on a 149-million-parameter ModernBERT backbone. The model is availabl

Models19.09.2026
Pocket FM unveils Sherpa, a narrative world model that keeps long-form stories consistent
Pocket FM has introduced Sherpa, a narrative memory system designed to solve one of the hardest problems in automated long-form storytelling: keeping

Models19.09.2026
AI nearly pulled the United States into a military operation against China
U.S. fighter jets were already airborne, heading for a Chinese vessel last spring, when intelligence driving the armed sortie turned out to be a chatb

Models19.09.2026
Researchers warn that slowing AI development remains an unsolved problem
Broad agreement among researchers that advanced models could become dangerous has not translated into a practical way to slow the pace. Raymond Dougla

Models19.09.2026
Chatbot-generated intelligence report nearly triggered military confrontation with China
A hallucination in an AI-assisted intelligence report sent the US military scrambling to intercept a Chinese vessel in the Middle East last spring, co

Models19.09.2026
Anthropic runs a wet lab in the Bay Area and uses its models to design and run physical experiments
Anthropic confirmed to TechCrunch that it operates a wet lab in the Bay Area where its models are used to plan and execute real-world biological exper

Models19.09.2026
Artificial Analysis publishes methodology for streaming speech-to-text leaderboard
Artificial Analysis has released the methodology behind its new leaderboard for streaming speech-to-text models. The headline metric, AA-WER Streaming

Models19.09.2026
Dario Amodei lays out plan to "Pace the Frontier," but how exactly would it work
A week after an Anthropic researcher published an apocalyptic warning that shook the industry, chief executive Dario Amodei put a more practical propo

Models18.09.2026
PrismML releases Ternary Bonsai 2 27B: 27B model at 5.9 GB, 98% performance
PrismML has shipped Ternary Bonsai 2 27B, a ternary-weight version of Qwen3.8 27B that occupies 5.93 GB against the 53.8 GB of the FP16 original. The

Models18.09.2026
AI super PACs pour nearly a million dollars into a sleepy South Dakota Senate race
The South Dakota Senate contest is not competitive. Polls show incumbent Mike Rounds leading independent Brian Beng by a wide margin after Democrat Ju

Models18.09.2026
Organizations report AI returns yet 62% lack storage readiness
A Seagate survey of 2,712 technology decision-makers at large enterprises reveals a stark gap: 99% of respondents expect AI to increase their storage

Models18.09.2026
One bug-reporting system to replace dozens of broken forms
Researchers who discover a flaw in a language model or deployed AI system currently face a maze: no single address, no uniform format, no guarantee th

Models18.09.2026
Epoch AI launches benchmark reviews: 9 of 15 fail
Epoch AI yesterday launched Benchmark Reviews, an initiative to audit the quality of the benchmarks the industry uses to measure language-model capabi

Models18.09.2026
Salesforce closes the gap between prototype and production with Agentforce
Salesforce is launching Agentforce, a platform built to solve the problem every enterprise hits: spinning up an impressive LLM demo in a weekend is ea

Models18.09.2026
New speech recognition model promises to end streaming text stutter
Most streaming ASR models spit out text quickly, but they tend to self-correct after the fact — words already on screen suddenly change when more audi

Models18.09.2026