OpenAI's GPT-6.1 Terra cuts errors by two-thirds at half the price
OpenAI's new mid-tier model replaces GPT-5.6 Terra in ChatGPT and Codex. It nearly matches Sol and Astra on errors and costs 50% less.
Model releases, benchmarks and capabilities. 131 stories, page 2 of 11.
OpenAI's new mid-tier model replaces GPT-5.6 Terra in ChatGPT and Codex. It nearly matches Sol and Astra on errors and costs 50% less.
BOSS Zhipin's AI lab released Nanbeige4.3-33B under Apache 2.0. It reports 80.9 on Terminal-Bench 2.1 and has a 512K context.
Moonshot's K3.1 Pro ranks 4th on the Artificial Analysis index at $2/$10 per 1M tokens. Open weights are due by Nov 4.
Anthropic's Claude Duet runs on Opus 5.5, watches your connected apps, sends a morning brief and asks before it acts. US Pro/Max waitlist.
Anthropic's new proactive agent Claude Duet can buy supplies through Stripe's Link and Shop Pay, and it asks before every payment. 1Password logins are next.
Meta released the weights of an 816B mixture-of-experts model with a 1M-token context. It is built for coding and agents and runs on two Mac Studios.
Google's new frontier model rose from 8th to 5th on Arena's agent leaderboard, still behind Opus 5.5, GPT-6.1 Sol, Fable 5.1 and GPT-6 Astra.
NVIDIA's new open multimodal model reads text, images, docs and video and claims 5x the throughput of open models its size.
Meta released open weights for Muse Spark Open, an 816B MoE coding and agent model with 42B active params and a 1M-token context.
OpenAI's Ultrafast tier for GPT-6.1 Sol is live. It's up to 8x faster than standard and costs $12/$60 per million tokens.
ChatGPT's new Intelligent UI turns some replies into interactive charts, calculators and diagrams. Paid users got it Oct 7; free users get it Oct 8.
Anthropic's new small model costs about 75% less to run than Haiku 4.5, and it's the first Haiku where you can set the effort level.