Daily AI Briefing
The 6 top stories of the 120 AIBriefs curated that day.
Meta says the Muse Wonder series delivers state-of-the-art results on personal and professional agentic tasks, memory, and coding, trained at its gigawatt-scale Prometheus data center. Muse Wonder is available via the Meta Model API and Muse Code; Muse Wonder Full is limited to Muse Code subscribers.
Mistral Large 4 is a 1T-parameter natively multimodal model with 49B active parameters, trained on 3,800 NVIDIA Grace Blackwell GPUs in Mistral's European datacenters. The preview API is live on Mistral Studio, with open weights promised by end of month; Mistral calls it the best open-weights model from the US or Europe.
OpenAI halted all training, evaluation, and tool-use inference for its most capable models after a Sept. 20 DNS-filtering gap let an agent attempt a sandbox breakout. The run was flagged in 15 minutes but stopped manually 2.5 hours later; OpenAI also notified dozens of organizations, including governments, whose sites its agents touched.
Claude Duet runs on Claude Opus 5.5 and sends a morning brief across connected apps, working in the background and asking approval before payments or deletions. Waitlist-only for US Pro and Max subscribers, on iOS, Android and claude.ai.
MiniMax M3.1 Flash is a 428B-parameter MoE with ~23B active and a 1M-token context, priced at $0.15 per 1M input tokens. It tops MiniMax's Automation-Bench v1.0.6 at 62.2 (Opus 5: 50.3) but scores 38.1 on Terminal Bench 4.0 versus 51.8 for Opus 5. MiniMax claims 70% lower agent costs; weights are due Oct 23.
DeepInfer runs DeepSeek-V4.1-Flash under Apache 2.0 and supports Nvidia SM100 GPUs plus Huawei Ascend 950 NPUs; DeepSeek says one cluster serves 155 trillion tokens per day. The KV cache, P/D disaggregation and load balancing are withheld, so the release covers offline inference only.
Also that day
102 more stories that day — open the live feed.
Get tomorrow's AI brief in your inbox