Daily AI Briefing
The 6 top stories of the 120 AIBriefs curated that day.
OpenAI's GPT-6 Astra adds advanced reasoning, computer use, and stronger writing and design judgment, positioned as its most intelligent and aligned model yet. Cognition's Devin uses it to test its own code, and Legora reviewed 41 documents in minutes, finding all four planted errors.
DeepSeek-V4.1-Flash ships with 552B backbone parameters, native image and text input, and up to a 1M-token context window. The YOCO architecture cuts GPU memory and prefill latency; Bloomberg reports pricing as low as a fraction of a cent per million tokens.
Fable 5.1 scores 55.8% on Terminal-Bench 4.0 in Claude Code, ahead of Fable 5 (42%) and Opus 5 (52.3%). Pricing matches Fable 5, but API cache reads drop 75% to $0.25/MTok; Perplexity ranked it first on its August WANDR eval at 0.601.
Meta Superintelligence Labs' fourth Muse Spark release in five months uses ~20% fewer tool calls and ~25% fewer tokens than Muse Spark 1.2. Pricing is $0.2 vs $25 per 1M output tokens versus Claude Opus 5, with a >90% discount if you opt in to training.
Two new GPT-Image-2.5 models: Sunburst for sharper details and precise edits, Flare for high-quality generation over 50% faster. Both support transparent backgrounds; rolling out to ChatGPT, ChatGPT Work, and Codex users.
JiT-DDT trains a text-to-image model with 3.6× fewer GPU-hours than the Linum v2 baseline while generating images with 4× the pixels. Code and weights are Apache 2.0 licensed; Linum calls it a research artifact, not a full model release.
Also that day
102 more stories that day — open the live feed.
Get tomorrow's AI brief in your inbox