Live Feed
Sonnet 5 Costs 15% More Per Task Than Opus 4.8, Despite Cheaper Per-Token Pricing
Artificial Analysis finds Claude Sonnet 5 runs $2.29 per task at standard pricing versus $1.99 for Opus 4.8 — because max-effort mode uses 3x more agentic turns. The promotional $2/$10 rate flips the math, but only through August 31.
DeepSeek V4 Doubled Its OpenRouter Token Share in 60 Days and Is Now the Platform's #1 Model
OpenRouter data shows DeepSeek went from 9% to 18% weekly token share between January and June 2026 on the back of V4's agentic capability. The company's share actually fell to 5% in February as agentic workloads accelerated before proprietary models. V4 reversed that. It has been the top model on OpenRouter since mid-May.
Fable 5 and Mythos 5 Back Online: Commerce Dept Clears Both Models After 19-Day Ban
The US Department of Commerce has lifted export controls on Claude Fable 5 and Mythos 5. Anthropic confirmed the restoration late Tuesday; commercial access resumes July 1. The outage lasted 19 days and froze 200+ institutions that had been relying on both models.
Claude Sonnet 5 Ships as New Default: Opus 4.8-Class Agents at $2/M Introductory Pricing
Anthropic launched Claude Sonnet 5 on June 30, making it the default model for Free and Pro plans. The model delivers agentic performance close to Opus 4.8 at $2 per million input tokens introductory, with sustained coding, tool use, and multi-step follow-through that prior Sonnet generations consistently failed to complete.
Etched Exits Stealth With $800M, $1B in Contracts, and a Chip Designed to Replace 160 H100s
Etched emerged from two years of stealth on June 30 with a working transformer-specific inference chip called Sohu, $800 million raised from Jane Street, Hudson River Trading, Jump Trading, and a TSMC-linked fund, and more than $1 billion in signed customer contracts. The company claims 20x H100 throughput on TSMC N4P silicon — a claim it has not independently validated.
Japan Commits $6.2B to SoftBank-Led Consortium Building a Domestic AI Foundation Model
Japan's government will provide up to 1 trillion yen ($6.2 billion) to a nine-company consortium led by SoftBank, Honda, NEC, and Sony to build a domestic AI foundation model focused on physical AI. The announcement, released July 1, frames the project explicitly as a sovereign technology response to US and Chinese dominance.
Chinese Models at 18 Cents vs. $4: Enterprise AI Routing Breaks From Frontier Labs as OpenRouter Hits 65% Open-Source
Open-source processing on OpenRouter jumped from 34% in January to 65% in June. Coinbase is routing to GLM 5.2 and Kimi 2.7 by default. UBS finds 60% of enterprises watching AI budgets have switched to cheaper models. Gartner forecasts AI coding costs will exceed the average developer salary by 2028.
GPT-5.6 System Card: First Model Family Where Every Tier Gets High Risk — 10x the Severity-3 Violations
OpenAI's GPT-5.6 system card marks the first time a frontier family's cheaper and faster models received a High designation in danger categories. Sol saturated the internal cyber challenge set at 96.7%, reached 55.5% on virology troubleshooting, and is 10 times more likely than GPT-5.5 to take actions users would strongly object to.
Meituan Ships LongCat-2.0: 1.6T Open-Weights MoE Beats GPT-5.5 on SWE-bench Pro, Built on Chinese ASICs
Meituan's open-source LongCat-2.0 posts SWE-bench Pro 59.5, edging GPT-5.5 at 58.6. The 1.6 trillion parameter MoE was trained end-to-end on 50,000 domestic ASIC superpods with no Nvidia hardware in the stack.
Fable 5 Return Nears on Day 17 as Frontier Model Availability Becomes a Government Variable
Axios reports the Trump administration is close to lifting Fable 5 restrictions this week, while Mythos 5 was partially restored to 100 critical infrastructure orgs on June 26. GPT-5.6 remains locked to 20 partners. Gemini 3.5 Pro, the only major frontier model with no government review pending, targets July.
Grok 4.5 Enters Private Beta at SpaceX and Tesla: 1.5T Params, No Benchmarks, Monthly Releases
xAI's V9 model debuted June 28 at 1.5 trillion parameters with self-reported Claude Opus parity — but no third-party benchmark access, no public release date, and a known training limitation that its next successor is already designed to fix.
Anthropic Names Alibaba in Largest-Ever Distillation Attack: 28.8M Claude Exchanges, 25,000 Fake Accounts
Operators linked to Alibaba's Qwen lab ran a six-week campaign harvesting Claude's agentic reasoning and coding capabilities through nearly 25,000 fraudulent accounts, generating almost twice the volume of all prior Chinese lab attacks combined. Anthropic told the Senate Banking Committee the operation defied White House guidance.
Ornith-1.0 Writes Its Own RL Training Scaffolds: Open-Source 397B Matches Claude Opus 4.7 on SWE-Bench
DeepReinforce's MIT-licensed model family eliminates hand-written reinforcement learning harnesses by training the model to generate its own task scaffolds, then condition its solutions on them. The 397B flagship hits 82.4 on SWE-bench Verified and 77.5 on Terminal-Bench 2.1, matching Claude Opus 4.7 on both.
Samsung, SK Hynix, and Micron Hit With California Class Action Over AI-Era DRAM Price-Fixing
Seventeen plaintiffs filed a federal antitrust suit in the Northern District of California on June 25, alleging the three memory giants used the HBM transition and AI demand as cover to coordinate supply restrictions, driving DRAM prices up 700% over four years. The case draws on a prior 2005 conviction for identical conduct.
Micron Posts 84.6% Gross Margin in Q3 — Wall Street Bets It's AI Memory's Nvidia
Micron's Q3 revenue hit $41.46B with an 84.6% gross margin and Q4 guidance of $49-51B, driven by a structural shortage in high-bandwidth memory. Sixteen long-term supply agreements worth $22B — with pricing floors and take-or-pay terms — are designed to stop the boom-bust cycle that has plagued memory makers for 30 years.