AI Daily · Nov 5, 2026: Kimi K3.1 Pro open weights ship with day-0 support across providers
Moonshot open-sources 2.8T-parameter Kimi K3.1 Pro weights with day-0 availability on Ollama Cloud, Nebius, SiliconFlow and Verda; OpenAI's agent pipeline patched 948 Linux kernel privilege-escalation bugs for $1.2M in API costs.
Showing the latest briefing (Nov 5, 2026). Every day is generated the next morning.
Pick a date
-
Moonshot open-sources Kimi K3.1 Pro weights with day-0 provider support
Moonshot released Kimi K3.1 Pro open weights: 2.8T total parameters, 104B activated, 1M-token context, targeting coding, 3D creation and financial research. It is available day-0 on Ollama Cloud (US-hosted, zero data retention), Nebius Token Factory, SiliconFlow and Verda Inference, and runs on 8x AMD MI355X via ATOM and vLLM with recipes and kernels published.
-
OpenAI agent pipeline patches 948 Linux kernel privilege-escalation flaws
OpenAI security researchers built an autonomous vulnerability discovery, exploitation and patching pipeline; in one run with an upcoming cyber model it validated and generated patches for 948 local privilege-escalation bugs in the Linux kernel at a total API cost of $1.2 million, using at least 10 previously unknown exploitation techniques. OpenAI says it is working with Linux maintainers and Patc
-
OpenAI launches ChatGPT Images 2.6 and two new API image models
ChatGPT Images 2.6 adds animated GIF output and reusable reference images, available today in ChatGPT and Codex across all tiers on web, desktop and mobile. Two new API models are also available: GPT-Image-2.6 Flare (images in as little as six seconds) and Sunburst (higher quality and more consistent edits), with image markup and SVG export.
-
Tencent releases Hy4 with vision, day-0 across platforms at 50% off
Tencent's Hy4 is out with vision for the first time plus gains in instruction following, long-context understanding and long-horizon agentic tasks, with a 1M-token context. Day-0 availability includes GMI Cloud (50% off for two months, unlimited), Vercel AI Gateway (50% off through Dec 31, 2026), OpenClaw, OpenCode Go and Chutes, reachable via Tencent TokenHub, TokenPlan and OpenRouter.
-
Stealth model Crouton Alpha free in Command Code with 1M context
Command Code is offering the new stealth model Crouton Alpha for free, featuring a 1M-token context window and multimodal reasoning, callable via the CLI with --model stealth/crouton-alpha.
-
Ling 3.1 Flash 1019 usage raised to $100 in Command Code
Command Code raised Ling 3.1 Flash 1019 usage from $10 GOAT credits to $100, with all higher plans including GOAT plus their extra usage.
-
Mistral Studio adds managed agents, apps and workflows
Mistral Studio now offers Apps and Managed Agents in public preview plus Workflows in GA on Mistral Cloud, reaching private cloud and on-premises deployments by end of November; models and the platform can run in the customer's own environment.
-
Ricursive releases Ripple-65B Preview dense open-weight model
Ricursive released Ripple-65B Preview, a 65B dense LLM pretrained from scratch, claiming state of the art among U.S. open-weight models on agentic coding and outperforming larger models such as Muse Spark Open-816B, Beam-501B, Inkling-975B and Nemotron 3 Ultra-550B. GMI Cloud is a Day 0 launch partner.
-
LiteLLM adds Kimi K3.1 Pro spend tracking and Bedrock GovCloud rates
LiteLLM's AI Gateway now tracks Kimi K3.1 Pro spending on TogetherAI, Fireworks and Baseten, and added Bedrock GovCloud rates for Kimi, Haiku and Grok plus regional and speed multipliers on custom per-call pricing. It also supports live speech translation via gpt-realtime-translate from OpenAI and Azure, billed per second of audio sent.
-
Tencent Hy4 scores 35 on AA Intelligence Index with 89% hallucination rate
Tencent's Hy4 scores 35 on the Artificial Analysis Intelligence Index v5, tying Grok 4.7 (xhigh) and trailing Claude Haiku 5.5 by one point, ranking second among China-based labs; but it shows an 89% hallucination rate on AA-Omniscience at roughly $0.78 per index task. In a separate side-by-side test, Hy4 rebuilt GTA6 trailer scenes in Blender for $12.34 versus $16.92 for DeepSeek V4.1 Flash 1017.