OpenAI Says Its Newest Models Got Far Better at COBOL
OpenAI reports big gains on its own tests for building and migrating code in COBOL, Fortran, OCaml and Visual Basic, with GPT-6.1 Sol hitting 86.6% on a COBOL migration eval.
#Models
6 stories
China Proposes Export Controls on 23 AI Technologies, From Large Models to Compute Clusters
China proposed adding 40 technologies, 23 of them in AI, to its export control list, exempting open-source tech and requiring exporters to register now.
DeepSeek Upgrades V4.1 Flash With Big Agent Gains and No Price Change
DeepSeek's V4.1-Flash-1017 update posts big jumps on coding and agent benchmarks, runs tasks faster, keeps the same API price, and ships open weights under the MIT license.
Reflection Releases Beam, a 501B-Parameter Open-Weight Model, Under Apache 2.0
Reflection put the weights for Beam, a 501-billion-parameter mixture-of-experts model built for coding and agents, on Hugging Face under the Apache 2.0 license.
Microsoft Completes Pre-Training of MAI-Base-2, a 2.8T-Parameter Model
Microsoft AI finished pre-training MAI-Base-2, a 2.8-trillion-parameter MoE foundation model it says is 4–12× more compute-efficient than DeepSeek's strongest open base model.
Meta Releases Muse Spark Open, an 816B Open-Weight Model That Runs on Two Mac Studios
Meta released the weights for Muse Spark Open, a multimodal coding and agent model it calls the most capable US open-weight model to date, built with Apple to run on two M5 Ultra Mac Studios.