GLM 5.3 Flash from Z.AI saw its input price drop 73% to $0.04 per 1M tokens, the largest single-day move across tracked models. The ecosystem gained 124,723 GitHub stars today across 10,000+ open-source AI projects, with 89 new projects entering the index. This week 1,850 new projects and 100 new Hugging Face spaces have been added. The top gainer was security-audit-skill with +3,617 stars.
AI News · November 11, 2026
GLM 5.3 Flash input price falls 73%
124,723 stars today89 new projects+3,617 security-audit-skill
Releases of the day
Browser Use 1.0.2 fixes MCP integration when Chrome remote debugging is off, letting the CLI drive its own Chromium. Open Interpreter rust-v0.0.61 ships without model-provider catalog changes. LiteLLM v1.109.0 arrives with 982 pull requests, 144 added features and 424 fixes. Phoenix CLI 1.22.0 moves to OpenTelemetry JS 3, updating the SDK trace and API dependencies. n8n 2.47.4 fixes expression transfers and observational memory. Milvus v3.0.6 and Puppyone v0.4.3 are also released with minor changes.
n8n [email protected] 2.47.4 (2026-11-11) · Bug Fixes · * core: Keep expression transfers working after built-in replacements (#43360) (b110abb) * core: Keep observational memory running after a...
Browser Use 1.0.2 Fixes on 1.0.1. · - MCP with Chrome's remote debugging off: browser-use --cli-mcp now drives its own Chromium instead of stopping at the chrome://inspect steps, and tells your...
OOpen Interpreter rust-v0.0.61 Open Interpreter 0.0.61 · Models · - No model-provider catalog or request-admission behavior changes are included in this release. · Release · - Attest Linux Open Interpreter...
LiteLLM v1.109.0 Verify this release · Same command for litellm-database, litellm-non_root and the docker.io/litellm/* images · What's new in 1.109.0 · 982 PRs. 144 added, 424 fixed, 60...
Milvus v3.0.6 Release note is coming soon...
Phoenix @arizeai/[email protected] Minor Changes · - 80bac51: Move @arizeai/phoenix-otel to OpenTelemetry JS 3 (@opentelemetry/sdk-trace 3.x, @opentelemetry/api 1.10) so spans created by apps on OpenTelemetry JS...
PPuppyone v0.4.3 Signed and notarized PuppyOne Desktop release. · Commit: ea82e6f7abf8da137d8676a571de4934b7bff05a Promoted Internal release: v0.4.3-internal.56 R2 version prefix:...
Pricing moves
Input prices rose for DeepSeek V4 Pro 0423 (+1% to $0.89), Llama 3.3 70B Instruct (+20% to $0.71), and GLM 5.3 Flash batch (+33% to $0.04). Prices fell for Qwen3 14B (-17% to $0.10), ReMM SLERP 13B (-14% to $0.30), Weaver alpha (-25% to $0.30), GLM 5.2 batch (-15% to $0.53), and GLM 5.3 Flash (-73% to $0.04). Users of Z.AI models see the largest swing.
▲ 1%DeepSeek V4 Pro 0423DeepSeek · input $0.88 → $0.89 per 1M tokens
▲ 20%Llama 3.3 70B InstructMeta · input $0.59 → $0.71 per 1M tokens
▼ 17%Qwen3 14BAlibaba · input $0.12 → $0.1 per 1M tokens
▼ 14%ReMM SLERP 13BUndi95 · input $0.35 → $0.3 per 1M tokens
▼ 25%Weaver (alpha)Mancer · input $0.4 → $0.3 per 1M tokens
▼ 15%GLM 5.2 (batch)Z.AI · input $0.62 → $0.53 per 1M tokens
▼ 73%GLM 5.3 FlashZ.AI · input $0.15 → $0.04 per 1M tokens
▲ 33%GLM 5.3 Flash (batch)Z.AI · input $0.03 → $0.04 per 1M tokens
Emerging projects
AnyModelBot (Python, 2,773 stars) offers an open-source, local-first alternative to Grok Bot and Claude Duet for running AI bots on any model. Kensaku (Rust, 1,502 stars) is a free file launcher with OCR and semantic search inside documents, code and screenshots. Project 0 (TypeScript, 1,079 stars) is an AI security harness with a web interface for code analysis and pentesting. Deepscient (JavaScript, 973 stars) aims to let AI tackle real scientific problems with 55 stages and 98 skills. RustXNG (Rust, 941 stars) is a self-hosted metasearch and grounding engine for AI agents. OpenVLHarness (Python, 440 stars) provides code for visual agentic tool-use with multimodal memory. Open-bot-avatar (TypeScript, 321 stars) defines a JSON spec for interactive agent avatars. Dori-superagent (TypeScript, 224 stars) is a proactive multi-session coding-agent harness.
0TypeScript · 1,079 stars · The open-source AI security harness. Includes a web interface for code analysis, pentesting, vulnerability...
open-bot-avatarTypeScript · 321 stars · A JSON spec and renderer for interactive agent avatars
dori-superagentTypeScript · 224 stars · A proactive multi-session coding-agent harness on herdr + omo: one lead, a duty loop, many lanes, a gated...
kensakuRust · 1,502 stars · Search inside documents, code, and even screenshots - instantly and locally. A free, open-source file...
DeepscientJavaScript · 973 stars · 让 AI 与人类一起攻克真正的科学难题 —— 12 阶段 55 门 · 36 工具 · 98 条技能 · 证据链与可复现
AnyModelBotPython · 2,773 stars · 🤖 Open-source, local-first alternative to Grok Bot, OpenAI dots and Claude Duet. Your own AI bots on any...
RustXNGRust · 941 stars · Self-hosted metasearch and grounding engine for AI agents, based on SearXNG. Rust, JSON API, optional page...
OpenVLHarnessPython · 440 stars · Official Code for OpenVLHarness: Real-World Visual Agentic Tool‑Use with Multimodal Memory
Research pick
A Hugging Face community paper titled "Don't Just Trust Your Memory" shows that external memory helps LLM agents learn from past tasks but can also cause memory-induced degradation, where agents fail on tasks they would normally solve. The proposed evidence-linked memory approach reduces this degradation.
The bigger picture
Today's facts point to continued price volatility among major model providers, a surge in new open-source projects focused on security and agent tooling, and ongoing research into the limitations of agent memory systems. The ecosystem remains highly active with steady project growth.
Source: olud.ai tracking of 10,000+ open-source AI projects, 300+ models and live provider pricing. Figures are measured, not estimated. All releases · Live pricing · Latest in AI
← All AI news