Ten instances of Anthropic's Claude Sonnet 5.5, run by AI benchmarking startup Vals AI in September, produced a 17,895-line Lean proof file, Solution.lean, establishing the pentagonal bipyramid as the…
Vals AI's Proof Bench found that most of the accuracy gain from increased reasoning effort occurs between the low and medium settings, with Opus 5.5 reaching 99% at medium effort for far less cost tha…
AI evaluation firm Vals AI reported that ten Claude Opus 5.5 agents, working over about 15 hours and posting 733 messages, produced a new shortest-path algorithm called C-HD with a machine-checked Lea…
Vals AI published a shortest-path algorithm, C-HD, devised by ten Claude Opus 5.5 agents under researcher Geby Jaff, along with a Lean proof of its correctness and runtime bound produced in about 15 h…
Vals AI, backed by Andreessen Horowitz, released CheatBench, a benchmark tool that measures how AI models cheat on different task categories. The release positions Vals as a neutral independent standa…
OpenAI launched Astra for Law, a legal-work product that combines the GPT-6 Astra model with a dedicated legal search index and legal-specific instructions rather than a separately trained base model.…
OpenAI Group PBC launched Astra for Law, a configuration of its GPT-6 Astra model wrapped in a legal search index covering more than 230 million URLs, which passed the overall correctness check on 54%…
OpenAI's GPT-6 Astra spent 141 hours in a Minecraft test run by evaluation firm Vals AI, progressing further than any AI system Vals AI had previously tested before a Creeper explosion destroyed its c…
Former OpenAI and Anthropic researcher Jacob Coxon left Anthropic, warning that leading AI laboratories are moving toward self-improving superintelligence before reliable controls are available. The d…
OpenAI's Astra reached a nether fortress in Vals AI's long-horizon Minecraft computer-use evaluation, a milestone Vals AI said no AI has achieved in history. Vals AI stated Astra is now in the final s…
A new analysis from Vals AI, an independent AI benchmarking company, found that longer, more complex AI tasks can have an environmental impact 10,000 times greater than simple queries, with building a…
Vals AI's Recursive Self-Improvement Index, launched in August 2026 with CoreWeave, shows the top-performing model, Claude Fable 5.1, scores 35.03% on autonomous AI development tasks. The index provid…
OpenAI released GPT-6 Astra on September 3, and Towards AI Deployment was named an OpenAI Select Partner. Astra scored 53 on Artificial Analysis's Intelligence Index 4.3, tied with Fable 5.1, and pass…
Anthropic's Claude Fable 5.1 decoded a 371-year-old royalist cipher, the 'Cyphral Distich' by Sir Thomas Urquhart, in 44 minutes without human help, according to Vals AI. The solution, which spells 'O…
Anthropic's new Claude Fable 5.1 model has taken the top spot on leading AI benchmarks, surpassing Chinese rivals, according to industry tests. The model also claimed first place on Vals AI's index fo…
Vals AI's Fable 5.1 model has successfully decrypted a 373-year-old cipher known as the Cyphral Distich, marking a milestone in AI's ability to handle complex, reasoning-heavy tasks. The breakthrough,…
OpenAI's GPT-5.2, released in August 2026, achieves 95.4% on SWE-bench Verified, surpassing Claude Opus 5 by 0.6 points and becoming the top open-weights model, according to the Vals AI leaderboard. T…
Vals AI has raised $40 million in a Series A round led by Andreessen Horowitz to build independent evaluation benchmarks for AI models, targeting the gap between standardized test performance and real…
DeepSeek's V4-Pro-0813 model scored 53 on the Artificial Analysis Intelligence Index, trailing OpenAI's GPT-5.6 Terra by 4 points and Moonshot AI's Kimi K3 by 7 points, while ranking 12th on the Vals …
XAI and Databricks announced a partnership on June 18 at the 2026 Data + AI Summit to integrate Grok models natively into Databricks' Agent Bricks platform, targeting enterprise document understanding…