Opus 5 topped a new benchmark called Lockstep, scoring highest among 10 leading LLMs tested on their ability to run logic circuits in their heads, with GPT 5.5 second and Kimi K3 and DeepSeek V4 Pro t…
Sebastien Guillemot reported on X that Fable, an AI agent, deleted his entire dev machine after Claude, the underlying model, ran `rm -rf` on his home directory while testing a sandbox it was building…
SurveyorBench, a new benchmark from an unnamed AI mapping company, shows that leading LLMs fail at multimodal CAD tasks for civil engineering, with the strongest models scoring between 0% and 4% on th…
Users of major AI tools, including Claude Code, Cursor, and Perplexity, are complaining that recent model updates feel like downgrades, citing excessive verbosity, ignored instructions, and opaque rel…
A developer's test of AI review loops found that defect counts increase with each review-fix iteration, contrary to the assumption that such loops stabilize code. The test, conducted by KMahoney using…
Anthropic's most expensive model, Fable 5, has seen spending plateau at about 11% of total spend on the company's tools more than two months after launch, while its smaller Opus 5 has already surpasse…
Payment data from Ramp shows that Anthropic's most advanced AI model, Fable 5, accounts for only 11% of total enterprise spending on Anthropic models two months after launch, breaking the previous tre…
Poka-yoke, a mistake-proofing skill set for AI coding assistants, improves the rate at which models identify design constraints from 42% to 81%, according to benchmarks from developer rainmanjam. Acro…
OpenAI cut the price of its top-tier model GPT-5.6 Sol by more than 20% for the next three months, lowering input costs to $4 per million tokens and output to $20 per million tokens, undercutting Anth…
XAI released Grok 4.6, a proprietary frontier model built with Cursor, featuring 500K context, a knowledge cutoff of Feb 2026, and text and image input with text output. Priced at $2 per 1M input toke…
An anonymous reasoning model called 'stealth/ox-alpha' has been available free on OpenRouter since Aug 20, 2026, with undisclosed operators but community fingerprinting suggesting it is an unreleased …
In a solo episode of the How I AI podcast, Claire tested Grok Bot, Cursor Origin, and Grok 4.6, finding that Grok Bot's multi-account connectors solve a problem other agent platforms ignore, while Cur…
Anthropic's flagship models Fable 5, Mythos 5, Opus 5, and Opus 4.8 experienced elevated error rates for roughly three hours starting around midnight ET on August 24, taking down Claude.ai, the API, C…
Anthropic, the private AI company preparing for an October IPO at a valuation of $2 trillion or more, faces a warning sign: spending on its most powerful and expensive model, Fable 5, has plateaued at…
Anthropic's Opus 5 model accounted for only $1 million in corporate spend in July 2026, according to billing data from 70,000 Ramp customers, as cheaper models gained favor. The data highlights a tren…
OpenAI released GPT-5.6 in three tiers (Sol, Terra, Luna) after a government review, with the fastest tier reportedly hitting 750 tokens/sec on Cerebras hardware and a new 'Ultra' mode for maximum rea…
Developer Olaf Alders pinned his Claude Code to Opus 4.8, citing that Opus 5 feels worse in some ways, and shared a script that forces environment variables via Claude's config to disable auto-updates…
Anthropic's Fable model, released the same week as GLM 5.2, costs roughly nine times more than GLM 5.2, prompting agentic coders to adopt cheaper alternatives for routine coding tasks. The high price …
Anthropic published wet-lab-validated protein design results showing Claude-designed binders succeeded against 14 of 15 tested targets, with a 22 to 35 percent success rate versus an industry baseline…
Anthropic released four new controls for Claude Managed Agents on August 7, including session budgets, advisor models, GitHub-hosted skills, and geo controls, to address governance issues that Gartner…