Oumi cofounder and CEO Manos Koukoumidis is building tools to help enterprises cut AI costs by using smaller, task-specific models instead of expensive frontier systems, a market that also includes co…
DeepSeek released V4-Flash on July 31st with an average cost of $0.03 per benchmark task, the lowest among well-known AI models tested by Artificial Analysis, according to Reuters. The model scored 50…
Alibaba released Qwen3.8-Max on August 2nd, a mixture-of-experts model with 2.4 trillion total parameters and 95 billion active parameters, available immediately through QwenCloud with open weights pr…
MiniMax has released two H3 base checkpoints on Hugging Face under a standard community license, but the full 2K workflow still requires hosted services. The 288 GB repository includes H3-Base-FL2VA a…
Alibaba Group Holding Ltd. released Qwen3.8-Max-Preview, a coding and office-work model, exclusively through its Alibaba Cloud Token Plan on July 19, with an August 3 promotional post calling it 'A Ne…
Arnav Gupta launched Prismor on July 6 as an open-source control plane that intercepts AI agent tool calls before execution, enforcing policies to allow, warn, or block actions. The tool, first releas…
CrowdReply, co-founded by Dawood Khan and Sheheryaar Khan, is pitching AI answer rankings as a $1M monthly growth channel, claiming in an X post that a startup launched the prior month can make $1M/mo…
OpenAI previewed Astra, an unreleased model built for long-running research tasks, by publishing 10 claimed advances in mathematics and theoretical computer science on August 1, with an internal versi…
Aikido Security identified the malicious Python package `anthropickit`, published on June 14, as a possible match for malware that Anthropic says its Claude Mythos 5 model created during a cyber evalu…
RuntimeWire, a one-person newsroom operated by founder Ryan Merket, has published 1,627 articles and recorded 205,000 pageviews since launching in May, with AI systems handling much of the reporting a…
Y Combinator's open-source QM agent runtime, version 0.1.4, has drawn 6,600 GitHub stars, 700 forks, 68 pull requests, and 14 open issues within two days of publication, according to a code review by …
Eigen Labs' lead research engineer Gajesh Naik said on August 1st that a public network of developers and AI agents made Poolside's Laguna XS 2.1 inference 138.7% faster on Apple Silicon, four days af…
DeepSeek-V4-Pro defeated Phi-4-reasoning 12 tasks to 0 with an aggregate score of 105.5 to 33.0 and 100% confidence in a head-to-head benchmark of 12 fresh text tasks scored by gpt-5.4. The evaluation…
Cyera, founded by Yotam Segev and Tamar Bar-Ilan, is pursuing Oasis Security through a letter of intent, a move CTech valued at about $1 billion, signaling a push to make AI-agent identity a full cybe…
Arrakis Security raised $8 million in a seed round led by Hetz Ventures on August 2, with participation from ElevenLabs CEO Mati Staniszewski, Torq CEO Ofer Smadari, Pentera CEO Amitai Ratzzon, and un…
Wafer, an AI infrastructure provider founded by Emilio Andere and Steven Arellano, reported that running Kimi K3 on eight AMD MI355X GPUs delivered 48 tokens per second per dollar, outperforming Nvidi…
Rodney Brooks, the roboticist who co-founded iRobot and now serves as co-founder and CTO of Robust.AI, argued in an August 1st essay that technology forecasts conflate four distinct time scales: resea…
CostPerPrompt has published a suite of online calculators that estimate monthly spending on AI workloads including chatbots, agents, retrieval systems, and voice applications, tracking pricing for mor…
Meta's Muse Spark 1.1 scored 51 on Artificial Analysis's Intelligence Index, tying with GPT-5.4, GPT-5.6 Luna, and GLM-5.2, and ranked third on SciCode, contradicting a Futurism article that claimed M…
Bernini-R Edit Video defeated Marey Realism V1.5 in a head-to-head video generation test, scoring 32.3 to 27.9 with a 94% confidence win, 3 task wins to 0, and one tie. The test, conducted by an unnam…