Daily AI Briefing
The 6 top stories of the 120 AIBriefs curated that day.
Get this briefing in your inbox every morning
Standard API prices are 50% below GPT-5.6 Terra, at $1/$5 per million input/output tokens. On Terminal-Bench 4.0 at max effort it solved 50.3% of tasks versus 21.5% for GPT-5.6 Terra, and low-effort factual errors fell from 53.6% to 16.7%.
OpenAI published model results on the Kakeya conjecture, the Exponential Time Hypothesis, and quantum chromodynamics, with external "shepherds" guiding the computer science and physics results. A group led by Russell Impagliazzo, Ramamohan Paturi, and Ryan Williams will examine the ETH disproof, which OpenAI posted with a Lean formalization.
Kimi K3.1 Pro is a 2.8T-total/104B-active MoE with a 1M-token context, scoring 55 on the Artificial Analysis Intelligence Index — 4th overall and ahead of Claude Fable 5.1, GPT-6 Astra and Gemini 4 Argon (all 53). Weights aren't out yet; Moonshot promises full release by Nov 4, 2026. API pricing is $2/M input and $10/M output.
Gemini 4 Argon scores 97.5% on ARC-AGI-3 and 91.3% on ARC-AGI-2, making Google DeepMind the second frontier lab above 95% on ARC-AGI-3. It also hits 87% on the Embodied Reasoning (ERQA) benchmark, and four Argon agents beat Minecraft in 27 hours with 2,783 self-written programs.
Muse Spark Open is an 816B mixture-of-experts model with 42B active parameters and a 1M token context window, scoring 39 on the Artificial Analysis Intelligence Index — the #1 US open-weights model. Meta trained it on NVIDIA GB300 NVL72 for agentic coding, and it adds audio and video perception to image and document understanding.
DeepSeek calls it a "minor update," but Terminal-Bench 4.0 jumps from 31.2 to 42.3 and DeepSWE hits 77.0, above Opus 5 (74.0) and GPT-5.6 Sol (73.0). It scores 45 on the Artificial Analysis Intelligence Index, up 6 points, and is the equal #2 open-weights model.
Also that day
102 more stories that day — open the live feed.
Get tomorrow's AI brief in your inbox