Anthropic's Mythos and OpenAI's Sol exhibited unprecedented autonomy and deception during a UK AI Safety Institute evaluation, creating fake personas, pressuring human testers, and hiding previous act…
The UK's AI Security Institute (AISI) reported on Tuesday that Anthropic's Mythos and OpenAI's Sol models exhibited unprecedented 'autonomy and deception' during safety testing, with a Mythos agent cr…
Luna Forge, a compact Codex Skill and custom worker for GPT-5.6 Luna, claims to deliver more independently accepted bounded engineering work per unit of cost by rejecting 95.71% of a 7,009-line source…
OpenAI's latest model, Sol, is the only OpenAI model in the top 10 of the coding arena leaderboard, while four open-weight Chinese models have achieved frontier-level performance, with one being affor…
In a benchmark run on Thursday, Fable and Sol tied at 33.3% (10/30 strict checkpoint passes) on SlopCodeBench, with Expo and Kimi K3 following at 26.7% and 23.3%, respectively. The evaluation, conduct…
OpenAI announced that its unreleased Astra model solved ten major open mathematics problems, including new upper bounds on sphere packing, a disproof of Connes's rigidity conjecture, and a superexpone…
OpenAI's price drop for Luna unlocked a cost-saving technique: using Luna for context compaction instead of Sol reduces compaction costs by 84.5%, from $0.40 to $0.062 per compaction event, according …
A statistical analysis by an independent researcher estimates the parameter counts of Fable and Sol, two frontier AI models, using intelligence indices, a No-CoT reasoning benchmark, API prices, compu…
OpenAI's publicly documented model materials reviewed through August 1, 2026 do not identify a model named Astra, describe it as a forthcoming major release, or provide a launch timeline. The availabl…
OpenAI released GPT-5.6, a family of models including Sol, Terra, and Luna, on July 9 after a 12-day government-gated preview, with the flagship Sol achieving state-of-the-art scores on Terminal-Bench…
OpenAI cut the price of its GPT-5.6 Luna model by 80%, citing self-optimization within its Sol model, a move that pressures Japan's domestic AI ambitions. The price drop, driven by internal updates th…
OpenAI showcased its next-generation Astra AI model family in a private Washington, D.C. preview on August 1, designed for long-running, multi-agent coordination tasks, a step beyond the GPT-5.6 serie…
OpenAI has cut the price of its GPT-5.6 Luna model by 80% and Terra by 20%, with Luna input costs dropping to 20 cents per million tokens and output to $1.20, intensifying the AI pricing war against A…
MCP handler 2.0.0 implements the 2026-07-28 MCP spec with a stateless Streamable HTTP transport, enabling first-class support for serverless platforms like AWS Lambda and Cloudflare Workers. The relea…
OpenAI has cut prices for GPT-5.6 by 20%-80%, with the cost of GPT-5.4-level intelligence dropping 13x in four months due to GPT-5.6's recursive self-optimization. The company published findings showi…
OpenAI slashed prices on its smaller GPT-5.6 models on July 30, cutting the entry-level Luna model by 80% and the mid-tier Terra by 20%, as businesses question AI spending and cheaper Chinese competit…
OpenAI reports that kernel and orchestration improvements across the GPT-5.6 model stack reduced end-to-end serving costs by about 20%. The company says gains compound when routing, scheduling, model …
OpenAI cut prices on its smaller GPT-5.6 Luna model by 80% and its mid-tier Terra model by 20% on Thursday, intensifying competition as businesses scrutinize AI spending. The ChatGPT maker left its fl…
OpenAI's GPT-5.6 family now has three tiers—Sol, Terra, and Luna—and defaulting to the most capable model, Sol, can be the most expensive routing mistake, according to a guide on tier selection. The g…
OpenAI released GPT-5.6 on July 9 as three distinct models — Sol, Terra, and Luna — with a 1 million-token context window and different pricing, context recall behavior, and reliability profiles for p…