AI Briefing#open-weights
The latest AI and developer news about #open-weights, with the original source and a short summary.
The latest AI and developer news about #open-weights, with the original source and a short summary.
Runway announced Praxis-1, a robot policy model pretrained on web videos, and previewed an open-weight release within months.
The upgraded weights rank #1 open-weight and #6 overall in Provider Voice, while also leading in Controlled Voice and Pronunciation Robustness.
The model supports up to 12 reference files, exports five separate audio stems, and ranks #1 among open-weights models and #7 overall on DesignArena’s text-to-video leaderboard.
The updated index features 60% private datasets and new benchmarks, with Claude Opus 5.5 scoring 58 compared to GPT-6 Astra's 57.
Meta reported 97 mentions of Muse and zero of Llama in its Q3 2026 earnings call, alongside a 2026 capex guide of $135B to $145B.
Ant Ling's upcoming 1019 checkpoint raises the Artificial Analysis Intelligence Index score from 41 to 44 and improves AutomationBench-AA performance to 68%.
The open-weights MoE model achieves a score of 45, significantly above the median of 19 for similar-sized models, while offering 230.4 tokens/sec output speed.
The model achieves top rankings in internal evaluations against larger open and closed competitors while running on consumer GPUs with as little as 8GB of VRAM.
The model achieves 250 ms median endpoint latency and integrates with Gemma 4 E4B and VoxCPM2 for voice agents.
The open-weight model reduces thinking tokens to roughly half of the base model's usage while maintaining valid tool calls in agent loops.
The open-weight Aplomb 1 model achieves a KL divergence of 0.123 and a Brier score of 0.065 on the Typed Decisions leaderboard.
The 7B model generates 2K images and supports natural-language edits using an accelerated 8-step denoising schedule.
The new open-weight model reduces flagged censorship responses from 89.8% to 2.8% while maintaining general benchmarks within ~1 point of the original.
Scheduled for release in October under the Apache 2.0 license, this sparse MoE architecture emphasizes inference compute efficiency.
The model scores 44.86 on Decision Index 0.2.1, ranking #1 among 4B models, and processes 1M-token documents in ~3 seconds via API.
The author contends that banning open-weight models while allowing closed models to advance increases the offense-defense cyber gap.
Kate Carruthers argues that true AI sovereignty for Australia lies in managed interdependence and the strategic use of smaller, open-weight models, rather than pursuing technological autarky or relying solely on proprietary frontier models.
Runway has announced Praxis-1, an open-weight world action model for robotics that leverages large-scale video pretraining to improve policy generalization, with a public release planned for the coming months.
Victoria achieves 70.0% on Terminal-Bench 2.1 using REAP pruning, while Maple improves Canadian regulatory accuracy from 6.6% to 21.8%.
The 550B-parameter model scored 535.4 out of 600 on the IOI 2026 problem set, surpassing the top human contestant's score of 498.27.
The 2B parameter Jeff model achieves an 83.1% score on a five-benchmark panel, slightly exceeding the published 83.0% score of the larger Jev model.
The new 280M-parameter draft model accelerates LFM2.5-VL-3B inference by up to 3.13x on device and 2.66x on H100 GPUs.
StepFun has released Step 5 Preview, offering intelligence on par with Kimi K3 at one-third the cost, with weights scheduled to open on October 15.
An analysis of involution and overseas expansion in the Chinese AI ecosystem, structural limitations, and implications for US developers.