AI News · November 19, 2026

Docling adds Office 2016 charts, GLM 5.2 price jumps 6100%

37,400 stars today87 new projects+2,509 dsh-our-free-model

Docling v2.151.0 shipped with support for Office 2016 charts including waterfall and treemap in pptx files, plus calendar invitation events in email parsing. The ecosystem gained 37,400 GitHub stars across 10,000+ tracked open-source AI projects today, with 87 new projects entering the index.

This week 2,335 new projects entered the index along with 100 new Hugging Face spaces. The top GitHub gainer today is "dsh-our-free-model" with 2,509 stars. Several minor releases landed across n8n, LiteLLM, CrewAI, LangGraph, Pydantic AI, and Phoenix.

Releases of the day

Docling v2.151.0 added reading waterfall, treemap and other Office 2016 charts in pptx files, and added calendar invitation events to email parsing. The fix category addressed a pptx issue. CrewAI 1.15.43 fixed detection of multimodal support for Claude 5. Pydantic AI v2.71.0 let Peers join an InMemoryPeerNetwork under the agent name and renamed PeerNetwork to PeerConnection. LangGraph 1.2.23 gave Send-dispatched tool calls the same state as other calls. Phoenix v21.2.1 added release notes and taught phoenix-harbor to build Harbor tasks and verifiers.

Pricing moves

GLM 5.2 from Z.AI saw input price surge 6100% to $0.62 per 1M tokens. Qwen3.5-9B from Alibaba rose 240% to $0.17 per 1M tokens. Qwen3.8 27B from Alibaba dropped 60% to $0.06 per 1M tokens. Kimi K2.7 Code from Moonshot AI fell 14% to $0.6 per 1M tokens. Qwen3 Next 80B A3B Instruct rose 11% to $0.1 per 1M tokens. Qwen3.5-27B rose 15% to $0.3 per 1M tokens. R1 0528 from DeepSeek rose 14% to $0.57 per 1M tokens. MiMo-V2.5-Pro from Xiaomi rose 9% to $0.48 per 1M tokens.

▲ 6100%GLM 5.2Z.AI · input $0.01 → $0.62 per 1M tokens
▼ 14%Kimi K2.7 CodeMoonshot AI · input $0.7 → $0.6 per 1M tokens
▲ 11%Qwen3 Next 80B A3B InstructAlibaba · input $0.09 → $0.1 per 1M tokens
▲ 240%Qwen3.5-9BAlibaba · input $0.05 → $0.17 per 1M tokens
▲ 15%Qwen3.5-27BAlibaba · input $0.26 → $0.3 per 1M tokens
▲ 14%R1 0528DeepSeek · input $0.5 → $0.57 per 1M tokens
▲ 9%MiMo-V2.5-ProXiaomi · input $0.44 → $0.48 per 1M tokens
▼ 60%Qwen3.8 27BAlibaba · input $0.15 → $0.06 per 1M tokens

Research pick

A Hugging Face community paper titled Tideway proposed phase-aware tiered memory management for MoE inference on consumer devices. The work addresses deploying large Mixture-of-Experts models locally on consumer PCs to balance data privacy and API costs.

The bigger picture

Today's facts show a market with sharp price corrections on some models and extreme spikes on others, while the ecosystem continues steady growth in projects and stars. The focus on local MoE deployment in research aligns with the broader push toward on-device AI.

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