Daily AI Briefing
The 6 top stories of the 120 AIBriefs curated that day.
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.
Meta Superintelligence Labs released the Muse Wonder series, its second-generation family, roughly 14 months after the lab was founded. Wonder is available via the Meta Model API and Muse Code; Wonder Full is limited to Muse Code subscribers. Cursor lists pricing at $4/M input and $20/M output tokens.
Sonnet 5.5 is the second model in the Claude 5.5 family after Opus 5.5, priced the same as Sonnet 5 at $2/$10 per million tokens with cache reads at $0.20. Anthropic says it runs 30%+ faster and costs up to 30% less for most work, with thinking on by default.
M3.1 Flash is a 428B-parameter mixture-of-experts model with ~23B active and a 1M-token context, priced at $0.15 per 1M input tokens. MiniMax says it cuts agent costs 70%; VentureBeat notes it beats DeepSeek and Opus 5 on a business benchmark but carries a high hallucination rate. Weights arrive Oct 23.
DeepSeek's update to V4.1 Flash scores 77.0 on DeepSWE (vs 74.0 for Opus 5, 73.0 for GPT-5.6 Sol) and jumps Terminal-Bench 4.0 from 31.2 to 42.3. Weights ship on Hugging Face under MIT; API runs $0.15/$0.60 per million tokens off-peak.
A 120B-parameter mixture-of-experts model activating 12B parameters per token, with up to 1M tokens of context and text, image, document and video input. Artificial Analysis scores it 26 on its Intelligence Index — the leading US open-weight model, double its predecessor Nemotron 3 Super (13) at ~60% fewer output tokens.
Also that day
102 more stories that day — open the live feed.
Get tomorrow's AI brief in your inbox