Anthropic just launched tools to show companies what their AI spending is buying
The frontier lab is lifting the curtain on how users spend money on its models ahead of a planned IPO.
News
Models & Labs
New models, releases and benchmarks from OpenAI, Anthropic, Google, Meta, Mistral and the open-weight world.
629 stories · Every headline links to the original publisher. HapnGo curates and groups; we do not copy articles.
Google Deepmind releases Gemini 4 Carbon just two weeks after Argon with gains across the board
Just two weeks after Gemini 4 Argon, Google DeepMind is releasing Gemini 4 Carbon, which tops the Artificial Analysis…
inkGround: Benchmarking Fine-grained Visual Grounding Capabilities of Vision-Language Models for Cursive Handwriting in the Wild
Human handwriting is produced as an ordered sequence of pen movements, but an offline image keeps only the final ink. Handwriting…
Google announces Gemini 4 Carbon, its latest frontier AI model that you can't use
Gemini 4 Carbon is essentially a slightly better version of Argon.
Google AI gets cagey when asked who won the 2020 elections
But the AI responses get more forthright when asked outside the US.
I'm building my own IDE using claude
Hi! I'm Daniel and I'm building Briiq, an AI-native IDE for designers. It works like a "Cursor for designers". For 5+ years I was…
Gemini 4 Carbon
AI summary Gemini 4 Carbon, a new AI model from Google, focuses on reasoning, cybersecurity and dependable autonomous task execution. It is reported to score 71.0% on CWE-bench and 88.2% on SEC-bench Pro, and 0.0% on Gray Swan's benchmark for indirect prompt injection. The model is said to flag missing information before acting and to carry out multi-step plans within project limits. Access is first going to cyber defenders via the Fairwind Program, with developers, businesses and consumers to follow later.
Gemini 4 Carbon: Our next frontier model
AI summary Google DeepMind unveiled Gemini 4 Carbon just two weeks after Gemini 4 Argon, claiming first place on the Artificial Analysis Intelligence Index 5.0 with 59 points. It also reports top security results: 88.2% on SEC-bench Pro, while its CWE-bench and ExploitGym marks were 71.0% and 54.8% respectively. A small set of cyber defenders gets access this week, followed by the Fairwind Program and then a broader rollout to developers and consumers. Costs are the same as Argon's, with input billed at $2 and output at $10 for every million tokens.
OpenAI teases Dots for Europe, but its Berlin DevDay was really about building a platform
AI summary At its DevDay Exchange 2026 event in Berlin, OpenAI said its assistant Dots should reach more European users within weeks, after discussions with regulators; personal Pro accounts in the EEA, Switzerland and the UK still lack access. The company reported that weekly active Codex users in Germany have risen twentyfold since January 2026, without giving absolute figures. OpenAI cited heavy resource needs and low demand for closing Sora in March and for not offering a video-understanding model. It also promoted the Agents API, plugins and reach to 1.2 billion weekly ChatGPT users to draw developers onto its platform.
Claude for Legal now verifies sources automatically, then hands the file to a lawyer
AI summary Anthropic has added pre-built interactive artifacts to Claude for Legal, part of Claude Enterprise, after requests from legal teams since the product's May debut. The tools help with jobs such as checking citations, reviewing contracts against internal rules, following draft legislation and comparing laws between countries, with the model tying each claim to its source before a lawyer checks the output. Over 60 integrations, Bloomberg Law among them, let users link their own databases, while firms including PwC and Harvey offer their own tools on the platform. Anthropic says built-in settings uphold data protection and information barriers between teams.
Nvidia's open Nemotron 3.5 Transcribe combines transcription, translation, and speaker recognition
Nvidia has released Nemotron 3.5 Transcribe, an open speech recognition model that handles transcription, speaker recognition…
New Artificial Analysis index widens the gap between open and closed models
AI summary Artificial Analysis released version 5 of its Intelligence Index, in which Kimi K3.1 Pro slid from fourth to ninth, trailing eight closed models. The firm now measures a 12-point gap between open models and the leaderboard leader, compared with 3 points before. Private test sets now make up 60 percent of the index to counter training on known benchmark questions, and Grok 4.7 lost 12 points under the change. Researcher Nathan Lambert argues that reinforcement learning infrastructure also explains the differences between labs.
Thin, Light and Built for Everything: NVIDIA RTX Spark Windows Laptops Gain New Models, Agent Tools and Games
NVIDIA RTX Spark powers thin and light laptops built for everything that developers, creators and gamers do. RTX Spark laptops…
Voice AI in vogue: Nvidia releases new open transcription model
Nvidia just released a free speech-to-text model that can transcribe up to 40 languages, but it only works on recordings, not…
xAI is buying out homeowners who promise not to sue over data center noise
A subsidiary of Elon Musk's xAI is offering to buy up homes in a Mississippi neighborhood, as long as residents don't join a…
‘Elonville’: US residents say Musk’s xAI offered to buy homes if they don’t sue over datacenter noise
Several Mississippi homes have been purchased near datacenters and one resident said realtor outlined target to buy more than 400…
With AI writing the code, entrepreneurial thinking is now what counts most for newcomers at Anthropic
Anthropic manager Felix Rieseberg says the classic coding interview is dead because Claude now solves small programming tasks…
The future of the firm
How AI might enter the org chart—beneath, beside, and between us.
Why hasn’t Google acquired OpenAI?
Given Google’s massive resources, AI expertise, and the strategic importance of ChatGPT, why hasn’t Google attempted to acquire…
Claude Opus 5.5 can now rebuild real programs from scratch — 18.5% on ProgramBench, up from 0% in May
According to Jack Clark's Import AI (Oct 26), Claude Opus 5.5 (xhigh) now resolves 18.5% of ProgramBench's 200 from-scratch…
Parents sue OpenAI, alleging GPT-4o coached their daughter off seizure medication
A family is suing OpenAI, saying ChatGPT coached their daughter to quietly taper off her seizure medication and keep it from her…
Free, daily
The HapnGo Briefing, in your inbox
One email a day with the AI stories that matter, why they matter, and one tool worth trying. No spam, one-click unsubscribe.