442 summarised stories about Large Language Models, each linking back to the original source. Browse all topics →
MarkTechPost 12 hours ago 21
DeepSeek released DeepSeek-V4.1-Flash-1017, a new checkpoint of its V4.1-Flash model that is live via its website, app, and API. It scored 77.0 on DeepSWE v1.1. DeepSeek also opened its Search and Fetch APIs to all developers, enabling cheaper agent coding plus a first-party web layer under one API key.
Latent Space 13 hours ago 18 ● 27 sources
Muse released its Muse Spark Open and Reflection Beam models alongside multiple frontier and decision-model updates covered in this AINews roundup. Muse Spark Open is an 816B-total, 42B-active MoE with weights quantized so experts-only fit to about 530GB. As a result, open-weight licensing, pricing, and benchmark rankings shift across cost, agent performance, and inference hardware/serving notes.
MarkTechPost 13 hours ago 24 ● 4 sources
TypeSafe launched Jev, a decision-model that outputs labels/scores or yes/no probabilities without generating text, and within about 4 weeks at least 20 other Jev alternatives were shipped by companies and developers. Cloudflare’s Clef (27B) scored 94.20 on BANKING77 F1 versus Jev’s 79.74 in Cloudflare’s own run. Developers now have many more hosted and self-hostable decision-model options, with differences in licenses, hardware support, pricing, and some models claiming Jev SDK/API compatibility.
MarkTechPost 14 hours ago 49 ● 2 sources
StepFun released Step 5 Preview open-weight checkpoints, including FP8 and BF16, on Hugging Face and ModelScope. The model has 600B total parameters with 27B active per token and supports up to a 1M-token context. Developers can now self-host it for agentic coding and long-document/finance workflows, with specified data-center GPU requirements (4x B300 for FP8 or 8x B300 for BF16) instead of using only the API.
SiliconANGLE 20 hours ago 43 ● 3 sources
Microsoft’s AI research team detailed its MAI-Base-2 base model, which is built to be converted into full LLMs via additional training. The model has 2 trillion parameters and activates 80 billion per prompt, and it was trained for 38 days on 16,000 Nvidia GB200 chips with 50 trillion tokens. Microsoft says MAI-Base-2 is up to 12 times more hardware-efficient, will be licensed to “trusted partners” on Azure, and will be used to power a new line of LLMs.
The daily briefing
TLDRocket reads all relevant sources, removes duplicate coverage, and summarises the day in two minutes. Follow companies and topics for alerts, or get the briefing in Slack. Free, no spam, unsubscribe anytime.
TLDRocket
17 October 2026
Replit’s break with Anthropic is the day’s clearest signal that “model preference” is becoming a business metric, not a philosophical stance. CEO Amjad Masad says Replit is ending its partnership and will default to OpenAI’s ChatGPT for Replit Agent, while still letting enterprises pick Anthropic models. Even more telling: enterprise revenue rose from 2% last year to 15% this year, and Masad frames ChatGPT as the easiest path to scaling that customer mix. At the same time, Anthropic is reporting its own internal shift—Claude-based agents took over more day-to-day work, and junior hiring fell sharply, with the share of junior staff among new technical hires dropping from 27% to 9%.
Read the full briefing →