Open-Source LLMs Close the Gap With Proprietary Models in 2026
Onyx AI's evaluation finds DeepSeek-V4-Pro leads coding, while Kimi K2.6 and GLM-5.2 top reasoning, making open-weight models viable primary tools.
4 Sep 2026
Google's 2026 AI Push: Verified Launches Amid a Verification Gap
Google's May 2026 I/O keynote confirmed major Gemini updates, agentic tools, and hardware. But an August blog post remains unverified, highlighting the risk of misreferenced AI announcements.
4 Sep 2026
New AI Evaluation Benchmark AIReg-Bench Is Inaccessible to Researchers
The OpenReview listing for AIReg-Bench cannot currently be retrieved, leaving methodology, metrics, and authors unverified at a time when AI evaluator benchmarks are gaining regulatory weight.
3 Sep 2026
Generative AI VC Funding Hits $407B, But OpenAI and Anthropic Take Half
A record half-year masks extreme concentration: OpenAI and Anthropic captured 53% of venture capital, while deal volume fell sharply. Sequoia and a16z lead the investor field.
3 Sep 2026
The Six Levels of Vehicle Autonomy, Explained
SAE's J3016 standard defines who is in control at each step from driver assistance to full automation. The crucial divide sits between Levels 2 and 3, where responsibility shifts from human to machine.
3 Sep 2026
Google Ships Gemini 3.7 Flash as Gemini 3.5 Pro Remains Missing
The new Gemini 3.7 Flash model arrives weeks after 3.6 Flash, signaling a focus on cost-efficient coding and agent models as the promised Pro tier remains unreleased.
2 Sep 2026
Qwen3.7 Max Tops Multilingual Benchmarks as Europe’s Sovereign Models Rise
Alibaba’s Qwen3.7 Max scores 100% weighted multilingual on the September 2026 BenchLM leaderboard, ahead of Claude Opus 4.5, while European and open-weight options reshape enterprise choices.
2 Sep 2026
Inside NVIDIA’s Tel Aviv Lab: The Research Engine Behind Real-Time AI
A prolific Israeli research team is tackling the hardest problems in perception, action, and reasoning—from training-free video matting to reliable reinforcement learning and safer language models.
2 Sep 2026
Text-to-Image Models Shift From Photorealism to Precision and Brand Control
The latest Text-to-Image Arena rankings show OpenAI, Microsoft, and SpaceXAI in a tight race, while new personalization methods like StyleDrop and Textual Inversion are changing what businesses actually need from generative image models.
2 Sep 2026
The Only AI Benchmarks That Still Matter in 2026
As traditional tests saturate, agent-based evaluations like Cybench and τ-bench are emerging as the new standard for measuring whether AI can do real work.
2 Sep 2026
Robotaxi Economics Are Finally Shifting—But Safety Blind Spots Persist
Falling hardware costs and NVIDIA-led alliances are making autonomous ride-hailing viable, yet incidents like the Zoox recall show emergency-scene perception remains a critical unresolved bottleneck.
2 Sep 2026
A National Standard Arrives for Self-Driving Cars
The U.S. government launches ASCEND, a three-year consortium to create the first national performance standards for autonomous vehicles, aiming to replace fragmented state rules.
2 Sep 2026Newsletter
Get the AI news that matters
One short brief with the day's most important AI stories — written for professionals.
We send a confirmation link. No spam. Unsubscribe anytime.