AI
OpenAI's GPT-6 Astra performed about 40% worse on safety research tasks than on comparable tasks, quietly withholding help and inventing fake obstacles. OpenAI can't fully rule out strategic motives.
Loading…
Search
No phones and 6 stories for “benchmarks”
AI
OpenAI's GPT-6 Astra performed about 40% worse on safety research tasks than on comparable tasks, quietly withholding help and inventing fake obstacles. OpenAI can't fully rule out strategic motives.
AI
Mistral has released the weights of Mistral Large 4 (Le Chonk), a 1T-parameter multimodal model trained in Europe that now leads open-weight coding benchmarks.
AI
Kimi K3.1 Pro scored 55 on the Artificial Analysis Intelligence Index, ranking fourth of 215 models and placing on the Pareto line for intelligence versus cost.
AI
A new Mercor study maps AI benchmarks onto 923 US occupations and finds only 10% of American working hours go to tasks those benchmarks actually test.
Tech
The DuckDB 2.0 alpha brings big speed gains on recursive queries, VARIANT storage and S3 reads. Here is what the benchmarks show and how to model data to benefit.
Phones
The Huawei Mate 90 Pro Max has surfaced on Geekbench with the Kirin 9050 Pro, an in-house nine-core chip that lands between the Snapdragon 8 Gen 1 and 8 Gen 2.