AI Digest

Category

Anthropic raises $81.3B in the largest IPO on record

The Rundown AI

A frontier AI lab is now publicly traded, so investors will scrutinize its pledge to slow down when safety requires it.

  • Anthropic began trading on the Nasdaq under the ticker ANT. Shares were priced at $83, above the target range, and closed their first day up 17% at $97.17.
  • The IPO beat the record SpaceX set five months earlier and values Anthropic at well over $1T. At the opening price of $111.50, the valuation briefly approached $2T.
  • CEO Dario Amodei said Anthropic will pace development until its models are safe and 'will not bend to either business nor political pressure.'
Read on nasdaq.com →

OpenAI releases GPT-6.2 Astra and GPT-6.2 Sol

TLDR AI

This is OpenAI's response to Anthropic's lead on the charts, but early reviews are mixed.

  • Astra is aimed at the hardest projects and Sol at everyday work. Both are available now in ChatGPT Work, Codex and the API.
  • On Artificial Analysis's ranking, Astra is ahead of Gemini 4 Carbon and behind only Claude Fable 5.2. Sol scores on par with September's GPT-6 Astra at about 15% of the cost per task.
  • OpenAI rated Astra at its 'High' threshold for AI self-improvement capabilities, a first for a model it has released publicly. Several testers criticized the models' coding and hard-to-read answers.
Read on openai.com →
Benchmarks
BenchmarkGPT-6.2 AstraClaude Fable 5.2Gemini 4 CarbonMuse WonderKimi K3.1 ProΔ vs best other
GPQA diamond96.896.096.5—94.2+0.3
FrontierMath-Tiers-1-3-v2-Private92.695.190.582.178.2-2.5
FrontierMath-Tier-4-v2-Private100.0100.095.170.765.9+0.0
OTIS Mock AIME 2024-2025100.099.7100.099.4100.0+0.0
SimpleQA Verified78.594.077.7—62.2-15.5
Chess Puzzles78.079.078.0—45.0-1.0
EBR-bench92.490.587.631.437.1+1.9
Mystery Game Puzzles95.090.073.051.042.0+5.0
MirrorCode67.8——45.643.3+22.2
ARC-AGI98.598.598.5——+0.0
SimpleBench79.294.992.6——-15.7
APEX-Agents72.280.184.266.165.5-12.0
ARC-AGI-295.893.894.2——+1.6
Furniture Assembly89.290.077.554.255.8-0.8
SpireClimb41.315.3——0.0+26.0

Δ = GPT-6.2 Astra minus the best other model listed. Sources: epoch.ai · Data: Epoch AI (CC BY 4.0)

Google launches Gemini 4 Neon, a fast model scoring near the top of the leaderboard

TLDR AI

Google sells Neon as a mid-tier model, yet it beats Google's own flagship on a major independent index, and Google has now shipped four Gemini 4 models in under two months.

  • Neon scores 61 on the Artificial Analysis Intelligence Index. That edges out Gemini 4 Carbon and trails only Claude Fable 5.2 and GPT-6.2 Astra.
  • In Google's own testing, Neon beat Claude Opus 5.5 and GPT-6.2 Sol on 16 of 19 benchmarks.
  • Neon is live for Pro and Ultra users. API pricing is $3 per million input tokens and $15 per million output tokens.
Read on blog.google →
Benchmarks
BenchmarkGemini 4 NeonClaude Fable 5.2GPT-6.2 AstraMuse WonderKimi K3.1 ProΔ vs best other
GPQA diamond96.396.096.8—94.2-0.5
FrontierMath-Tiers-1-3-v2-Private91.295.192.682.178.2-3.9
FrontierMath-Tier-4-v2-Private97.6100.0100.070.765.9-2.4
OTIS Mock AIME 2024-2025100.099.7100.099.4100.0+0.0
SimpleQA Verified75.294.078.5—62.2-18.8
Chess Puzzles83.079.078.0—45.0+4.0
Mystery Game Puzzles85.090.095.051.042.0-10.0
SimpleBench94.494.979.2——-0.5
APEX-Agents82.980.172.266.165.5+2.8
Furniture Assembly89.290.089.254.255.8-0.8

Δ = Gemini 4 Neon minus the best other model listed. Sources: epoch.ai · Data: Epoch AI (CC BY 4.0)

Musk, Nvidia staff and investors push to bring back Jensen Huang as he reportedly weighs a return

The Rundown AI

Pressure from employees and investors could reverse the leadership change at the world's most important AI chipmaker.

  • A petition to reinstate Huang passed 140K signatures within days, and more than 7,000 signers say they are Nvidia staff. Elon Musk signed it too.
  • The Financial Times reports that Huang is privately weighing a return, and that an ally is building a 'war room' of tech executives to lobby for it.
  • Nvidia's board has not responded publicly and is still choosing a search firm.
Read on businessinsider.com →

Hugging Face switches Inference Providers to OpenRouter's router for language models

Hugging Face Blog

Developers can now call any open-weight model on the Hub that a provider serves using one Hugging Face token, with automatic routing and failover between providers.

  • Language-model requests now go through OpenRouter. It picks a provider on live price, speed and reliability, and if a provider fails it retries on another one within the same call.
  • The router supports OpenAI's Chat Completions and Responses APIs and Anthropic's Messages API. Hugging Face adds no markup to provider prices.
  • Breaking change: custom provider API keys no longer work for language models.
Read on huggingface.co →

OpenAI reportedly folds its robotics team into core research, but staff dispute the report

The Rundown AI

The report suggests OpenAI thinks general models can now control robots directly, without a separate specialized model.

  • According to Fortune, OpenAI moved the robotics team into its main research and product groups and said models like GPT-6 Astra can run robots without a separate brain.
  • Engineer David Schnurr called OpenAI robotics 'alive and well' and pointed to 15 open roles. Another staffer called the report 'completely wrong.'
Read on fortune.com →

Google opens Genie 4 interactive world generation to AI Ultra subscribers

TLDR AI

Users can now explore AI-generated worlds and edit them in real time.

  • Eligible Google AI Ultra subscribers can use Genie 4 in Project Genie.
  • Users can edit a world by typing prompts while they explore it.
Read on blog.google →

Gemini for Mac can now use your computer in a background window

TLDR AI

The agent works across your local apps without taking over your screen.

  • Gemini handles multi-step tasks across local files, desktop apps and the web in an isolated window, with a live preview in the chat.
  • Users choose which apps Gemini can access, can pause it at any time, and approve key actions before they happen.
Read on blog.google →

Nvidia announces Rubin HPC, a GPU for the simulations AI agents run

TLDR AI

Nvidia is building chips for scientific and engineering simulation workloads driven by agents.

  • It has more than 50,000 cores, a record for a single GPU.
  • It is expected to be available in 2028.
Read on blogs.nvidia.com →

Alibaba releases Qwen4-Max, aimed at 3D, game development and professional work

TLDR AI

Qwen's new flagship moves into agentic creative and professional tasks, and smaller models in the family are due to be open-sourced.

  • Qwen4-Max can use Blender for geometry modeling, scene assembly, materials, lighting and rendering.
  • It works with leading game engines and can deliver a playable game, including gameplay logic, HUD design and visual effects.
  • Smaller models in the family are coming to open source.
Read on qwen.ai →
Benchmarks
BenchmarkQwen4-MaxQwen3.8-MaxΔ vs best other
Work360 (industry tasks)52.747.7+4.9
Last Secure Code Bench - function-level pass rate (works and is secure) (%)31.822.0+9.8
Last Secure Code Bench - secure code at function, file and repository level (%)14.78.0+6.7

Δ = Qwen4-Max minus the best other model listed. Sources: qwen.ai

OpenAI launches ChatGPT Wallet so agents can buy things without seeing card details

TLDR AI

Agentic commerce gets a payment layer from a frontier lab, backed by the major card networks.

  • AI agents can make purchases without access to the user's actual card details.
  • Every transaction requires a security check.
  • The Rundown reports it was built with American Express, Mastercard and Visa.
Read on openai.com →

Anthropic updates Projects in Claude Code with a more proactive coordinator

TLDR AI

The update makes it easier to manage several parallel coding threads from one project chat.

  • The coordinator is now more proactive.
  • It can answer questions in the project chat and give status updates on ongoing threads.
Read on threadreaderapp.com →

OpenAI models cut chip circuit area by 10–13% in its Jalapeño AI chip

TLDR AI

Frontier models are now contributing to the design of the hardware that trains them.

  • Design changes proposed by OpenAI's models cut the circuit area of key parts of Jalapeño by an estimated 10% to 13%.
  • Working with AMD, the same approach delivered up to 14% improvements on AMD's flagship GPU designs.
Read on openai.com →

Google's Mirror Dimension tests browser agents without touching the live web

TLDR AI

A controlled environment makes it safer to evaluate how secure and effective browser agents are.

  • Agents are tested in a controlled copy of the environment instead of on real websites.
  • Google says it is meant to evaluate both security and effectiveness while reducing real-world risk.
Read on blog.google →

Google's ZERO-TIM gets bitwise-identical logprobs between sampler and trainer for on-policy RL on TPUs

TLDR AI

Removing numerical drift between the serving and training paths keeps RL training truly on-policy.

  • It was verified across 64 TPU v5p chips.
  • It uses consistent arithmetic and standardized reduction orders to remove floating-point addition differences between the two paths.
Read on developers.googleblog.com →

Physicist uses GPT-6 Astra to build a candidate lattice formulation of the Standard Model

TLDR AI

An example of a frontier model contributing to a long-standing problem in fundamental physics.

  • Nathaniel Craig worked with GPT-6 Astra on defining the Standard Model without relying on perturbative approximations.
  • Astra helped build the mathematical framework and develop a detailed proof supporting the approach.
Read on openai.com →

Anthropic explains how to build strong safety evaluations

TLDR AI

Safety evaluations shape deployment and training decisions at frontier labs.

  • Anthropic stresses the importance of robust safety evaluations for AI models.
  • It says these evaluations inform deployment and training decisions.
Read on anthropic.com →

Thinking Machines examines end-to-end 4-bit language model training

TLDR AI

Training in 4-bit precision could sharply cut training costs if the accuracy problems can be managed.

  • The post covers quantization errors and how they affect model accuracy.
  • It compares the BF16, MXFP8 and NVFP4 formats.
Read on thinkingmachines.ai →
BenchmarkGPT-6.2 AstraGemini 4 NeonGPT-6.1 SolInklingQwen4-MaxClaude Fable 5.2Muse WonderKimi K3.1 ProDeepSeek V4.1 Flash 1017
GPQA diamond96.896.395.488.3—96.0—94.293.2
FrontierMath-Tiers-1-3-v2-Private92.691.293.733.3—95.182.178.271.2
FrontierMath-Tier-4-v2-Private100.097.6100.04.9—100.070.765.946.3
OTIS Mock AIME 2024-2025100.0100.0100.088.9—99.799.4100.099.4
SimpleQA Verified78.575.273.940.3—94.0—62.242.5
Chess Puzzles78.083.068.021.0—79.0—45.046.0
EBR-bench92.4—54.3——90.531.437.1—
Mystery Game Puzzles95.085.080.0——90.051.042.049.0
MirrorCode67.8—61.1———45.643.3—
ARC-AGI98.5—98.579.5—98.5——94.5
SimpleBench79.294.482.950.0—94.9———
APEX-Agents72.282.960.033.8—80.166.165.569.8
ARC-AGI-295.8—94.236.5—93.8——78.1
Furniture Assembly89.289.280.0——90.054.255.8—
SpireClimb41.3—23.8——15.3—0.0—
PostTrainBench——38.8——44.624.2——
ProofBench——99.00.0—100.084.091.0—
FrontierCode——50.214.0—————
LMCA——62.837.6—70.656.857.6—
DTBench——96.587.5—99.295.293.3—
WeirdML v3——50.64.4—65.324.918.919.7
WeirdML———32.3—————
FrontierSWE———4.1—69.9———

Highlighted: released in this period. Bold: best score. Independent scores from Epoch AI (CC BY 4.0).