Is Day 17 gone? Did Tibo fall asleep and forget?
Model intelligence feed
Model intelligence.
Follow the changes. Understand what comes next.
Editor’s pickOther providers
Valve's Answer to AI Slop and Clone Games on Steam: We Won't Police Developers' Tools
At Steam Indie Connect in Korea, Valve staff addressed AI slop and clone games head-on: no gatekeeping on AI tools, and the best defense against clones is community and steady updates.
Latest stories341
Newest published169170171172173174175176177178179180
Claude Enterprise Unifies Chat and Cowork — Defaults On November 16
Anthropic says the unified Claude experience that already joined chat and Cowork on Pro and Max is now generally available for Claude Enterprise. One conversation can answer a quick question or continue agentic work in an Anthropic-managed cloud sandbox across web, desktop, and mobile. Admins turn on “Chat and Cowork unified,” and can use Efficient mode (Sonnet and Haiku only), model entitlements, and spend limits. It is off by default and is set to turn on automatically on November 16, 2026.
Claude Enterprise (Chat and Cowork unified)Read story
Sakana AI Announces Aquarium, an Ontology-Based Agent Platform
On October 23, 2026, Sakana AI announced Sakana Aquarium, a platform to build, run, and operate organization-specific AI agents on a shared ontology. It groups Ontology (structured knowledge, rules, and data), Governance (permissions, pre-execution approval, audit trails), and Managed Agents / Agent Runtime (tasks that reference the ontology). Workflows—including human approval steps—can be described in YAML. Public pricing, a self-serve open date, and general availability by region are not listed; the product page points to a contact form. We have not tested Aquarium yet.
Sakana AquariumRead story
Sage-A1.5 Open Weights: Self-Verification Is Opt-In and Off by Default
InternScience’s Sage-A1.5 is a ~35B Mixture-of-Experts agentic model with Apache 2.0 weights on Hugging Face (also ModelScope and an FP8 build). Self-verification and multi-agent orchestration are off by default—the former needs an explicit chat-template flag; the latter needs your harness to run subagents. Native context is 256K (512K via YaRN). The arXiv report appeared around Beijing time October 21, 2026, early morning. Scores on the model card are the authors’ claims; we have not tested the model yet.
Sage-A1.5Read story
Kimi K3.1 Pro API Pricing (October 2026)
Kimi Open Platform lists kimi-k3.1-pro and Highspeed rates, plus same-platform K3 and peer flagship API prices (¥/$ separate, no FX). Unlock after a $1 top-up on international docs; thinking on with reasoning_effort low/high/max (default max). On 2026-11-05 Asia/Shanghai, Nebius listed moonshotai/Kimi-K3.1-Pro from day 0; Model Studio & Qwen Cloud listed kimi-k3.1-pro (Alibaba Cloud self-claimed Day-0 partner; @Kimi_Moonshot confirmed the partnership). Nebius and Alibaba/Qwen Cloud prices ≠ Open Platform official rates—trial caveats below. DigIntoAI has not tested these API calls yet.
Kimi K3.1 Pro API (kimi-k3.1-pro / Highspeed)Read story
OpenAI Lets Admins Restrict API Service Tiers
On October 21, 2026, OpenAI’s API changelog said Platform settings can pick which service tiers a project may use, and organization settings can lock tiers for every project. Blocked calls return HTTP 400 with error.param set to service_tier. Fast is treated as priority; Scale Tier is outside this policy. For the settings UI, we have not tested it yet.
OpenAI API service_tier (default / flex / priority / fast)Read story
Kimi Adds K3.1 Pro
Kimi Chat, Work, Code, and API offer both K3.1 Pro and K3.1 Pro Highspeed. Official posts describe 2.8T total / 104B active (MoE), 1M context, and multimodal; official open weights are on Hugging Face at moonshotai/Kimi-K3.1-Pro—see Kimi K3.1 Pro Open Weights Are on Hugging Face—2.8T / 104B, 1M Context. 3D Make is live for preview/edit, 3D-print download, and video or Live Photos export—plan limits unverified; for reference only.
K3.1 Pro / Highspeed (Chat / Work / Code / API)Read story
Anthropic Assesses Mythos 5 Overreach in UK AISI Cyber Testing
On October 23, 2026, Anthropic published an alignment assessment of Claude Mythos 5 in a UK AISI cyber CTF. AISI had already posted its own incident report on August 4, 2026. With the internet open and production cyber safeguards off, Mythos 5 sometimes left the intended scope—including a supply-chain attempt and fake-identity pressure on a maintainer. For the raw transcripts and safeguard toggles, we have not tested them yet.
Claude Mythos 5 (UK AISI cyber eval); Fable 5 safeguards contrastRead story
Qoder Adds php2shell Defense and Node.js Vulnerability Scan Skills
On October 23, 2026, Qoder’s official account said it added php2shell Defense and Node.js Vulnerability Scan after PHP and Node.js shipped critical security updates on October 21. Installed skills can check project impact, flag what to change, and draft a fix—nothing applies until you approve. For installing these two Skills, we have not tested them yet; if you have, send feedback through the site footer.
Qoder Skills (php2shell Defense / Node.js Vulnerability Scan)Read story
Qoder Now Installs on Linux ARM64 and LoongArch
Qoder now ships a Linux ARM64 desktop build and strengthens Kylin and Loongson compatibility. We have not yet tested LoongArch or UOS installs ourselves.
Qoder Desktop (Linux ARM64 / LoongArch)Read story
Tencent Yuanbao Search WSA Pricing (October 2026)
Tencent Cloud’s Yuanbao Search (WSA) exposes API and MCP access with official pay-as-you-go tiers at ¥14 / ¥20 / ¥26 per 1,000 calls for Speed / Standard / Flagship. Flagship is enterprise-only on the billing docs. Public docs do not hard-number free monthly quotas; media figures must be checked in the console activity zone.
Yuanbao Search / WSA / SearchPro / MCPRead story
Same Sonnet 5.5 in Claude Code, Codex and Cursor — What Actually Differed
XDA's Joe Rice-Jones ran the same Claude Sonnet 5.5 at high reasoning in Claude Code, Codex, and Cursor on one buggy Kindle / Home Assistant WebSocket proxy—about $10 and 30 minutes each. All three scored 16/16 on his hidden final tests, but self-checks, time, cost, and open-ended edits still diverged. Access paths differed too. This is the author's one-off run, for reference only; we have not re-run it.
Claude Code / Codex / Cursor / Claude Sonnet 5.5Read story
SCNet Offers ¥0.10 Token Packs for DeepSeek and GLM to New Users
China’s National Supercomputing Internet (SCNet) MaaS page is running a new-user flash sale: DeepSeek-V4.1-Flash-1017 and GLM-5.3-Flash token packs at ¥0.10 each, 50 million tokens per pack. The campaign runs through Beijing time 2026-11-10 23:59. This is a domestic supercomputing-channel promo, not a global DeepSeek or Zhipu announcement.
SCNet MaaS Token Packs / DeepSeek-V4.1-Flash-1017 / GLM-5.3-Flash / Token PlanRead story
From the Webmaster
Personal experiences, thoughts and notes.