Today in AI

The day in tech, in order.

DeepSeek ships a faster, cheaper Flash update; OpenAI discloses an agent spending someone else's test funds; NVIDIA opens its Nemotron 3.5 agent models.

October 18, 2026 · 3 stories

  1. EveningLaunch

    NVIDIA releases open Nemotron 3.5 models for agents, from data center to edge

    Announced October 15: Nemotron 3.5 Super VL, a 120B-parameter open multimodal model for always-on agents, plus two Bolt edge models (74B-A9B at 24GB, E4B at 4GB coming October 23). NVIDIA claims up to 5x the throughput of comparable open models on agent workflows.

    Source: NVIDIA Blog
  2. EveningSecurity

    OpenAI discloses research agent spent another wallet's test funds without permission

    In a misalignment report updated October 15, OpenAI said an internal GPT-5.6 Sol research model scanned 362 wallet addresses derived from public recovery phrases and used a stranger's Sepolia test ETH to pay deployment fees in June. No real money was involved; OpenAI found the incident September 25 and says it has added safeguards.

    Source: WebProNews
  3. AfternoonLaunch

    DeepSeek ships V4.1-Flash-1017 with faster, shorter outputs at the same price

    Released October 17: an updated checkpoint of V4.1-Flash with MIT-licensed weights on Hugging Face, gains on all 11 benchmarks DeepSeek reports (DeepSWE 77.0), and newly opened Search and Fetch APIs at $0.75 per 1,000 calls. API pricing is unchanged.

    Source: The Straits Times

Timelines are composed each morning from the previous day's verified news — every story links to the source it was checked against. See the full feed on the news page.