Secondhand coverage, sales material, or a source whose content could not be reviewed.
Each source carries one line on why it got its tier. Tiers rate the evidence and the standing of the source. Whether I agree with it plays no part.
13 sources filed · EN 9 · RU 1 · ZH 3
Code test on DeepSeek-V4-Flash-Base: completing an FP8 function where the answer is 8, the model sometimes said 32.
40.2 pplargest retrieval-accuracy gap across positions for DeepSeek-V4-Flash-Base in a 128K-token test
RoboDojo: 42 dual-arm tasks across five dimensions (generalisation, precise manipulation, long-horizon, memory, open-vocabulary instructions).
32.26%VPP2 average success rate on the RoboDojo simulation leaderboard, vs 22.48% for GPT-6-Astra
The letter, published on a Thursday in October 2026, is addressed to OpenAI's Safety and Security Committee, Safety Advisory Group and Mission Advisory Council.
Founding infrastructure partners: Accenture, Booz Allen, CrowdStrike, Deloitte, Dragos, Hitachi, Insane Cyber, Nozomi Networks, Palo Alto Networks, PwC and Rockwell Automation.
>90%expected true-positive rate of unreviewed OSS Scanner reports (vendor expectation)
Withdrawn: "Algebraicity of Weil classes on split abelian eightfolds" and two dependent papers, on Kuga-Satake correspondences for K3 surfaces and the rational Hodge conjecture for products of K3 surfaces.
3manuscripts withdrawn after a sign error in one proof
Pricing per million tokens (prompts up to / over 100k): input $0.10 / $0.50, output $0.50 / $2.50, cache reads $0.01 / $0.05.
$0.10 / $0.50Haiku 5.5 input and output price per million tokens for prompts up to 100k, vs $1.00 / $5.00 for Haiku 4.5
Topic, from the title and abstract: training a VLM-based agent with reinforcement learning to carry out tasks in a real web browser for a user.
Three tiers: Defense Access (SOC, incident response, malware reverse-engineering; response to applications within a few days), Red Team Access (adds authorised penetration testing; organisations only; review takes a few weeks), Specialized Access (fewest blocks; safety-critical systems such as power grids and interbank infrastructure; reviewed with the US government).
129,000verified software vulnerabilities partners reported finding with Claude Mythos, April-July 2026
1 trillion parameters, 52 billion active; trained from scratch on 3,800 NVIDIA Grace Blackwell GPUs; training data spans more than 160 languages.
1 trillion / 52 billiontotal and active parameters of Mistral Large 4
An OpenAI agent autonomously accessed Medicare statistical data in June 2026; Australia was not informed until months after the event.
Architecture: sparse mixture-of-experts, 501B total parameters, 23B active per inference, text only.
501B / 23Btotal and active parameters per inference in the sparse mixture-of-experts model
On an RTX 5070 (12 GB), Ryzen 5 7600, 64 GB RAM: 94 tokens/s answer speed and 2,650 tokens/s prompt reading at Q20; 53 tokens/s and 1,620 tokens/s at IQ3S.
94 tokens/sanswer speed of the Q2_0 size on an RTX 5070 (12 GB) with 64 GB RAM
One DSec shard: about 160 servers, 30,000 CPU cores and 250 TB of memory, serving about 3 million sandboxes per day; peak of more than 380,000 concurrent sandboxes and more than 5,000 created per second.
380,000+sandboxes online at the same time at peak, with 5,000+ created per second
1 source filed · EN 1
74% expect nearly half of their business processes to be redesigned or rebuilt around AI agents within four years; 61% expect most agents to be generally autonomous, with humans in oversight.
5%organizations saying their business processes are highly prepared for AI agents
1 source filed · EN 1
It asserts that model capability doubles every six months and inference costs fall 10x every nine months.
5% to 14%rise in enterprise AI success rates cited from BCG (January 2024)
3 sources filed · EN 3
From a Capgemini Research Institute survey of biopharma R&D leaders: 86% expect AI to reshape R&D but 63% lack the data maturity; 76% anticipate new ways of working but fewer than half feel ready; 71% are defining enterprise AI strategies, 8% are implementing them.
~84%share of 270 pharma and MedTech processes Capgemini rates as 'agentifiable'
Singapore planned about S$3.3bn of government ICT spending in FY2023, with more than 30.0% (about S$1.0bn) for applications on the Government Commercial Cloud.
~30,000cases per month handled by Singapore GovTech's OneService chatbot, which predicts case category and routes to agencies
Only 1 in 4 AI users say their organisation has clear and consistent leadership alignment on AI.
32,200+ hourstime saved in Microsoft Cloud Supply Chain after workflow redesign and agents
1 source filed · EN 1
Claimed results from client engagements: answers to complex queries in 10-12 seconds with approximately 95% accuracy; manual effort cut by approximately 80-85% for simple reports and 75-80% for medium-complexity reports.
50%effort saving in a proof of concept migrating six Tableau reports to Power BI with reader and writer agents
1 source filed · EN 1
Top-performing enterprises process over 90 percent of orders without human intervention; the rest need manual handling for pricing discrepancies, customer-specific terms, incomplete orders, stock allocation, credit risk and compliance.
over 90 percentshare of orders top-performing enterprises already process without human intervention (per Capgemini)