AI News · August 19, 2026

DeepSeek V4 Flash and GLM 5.2 lead a broad price surge across AI models

149,514 stars today51 new projects+5,661 public-apis
By olud.ai editorial · built from our own tracking data, published daily

Multiple AI model providers raised prices today, with GLM 5.2 input cost jumping 159 percent to $1.19 per 1M tokens. DeepSeek V4 Flash input price rose 14 percent to $0.08 per 1M tokens. The ecosystem gained 149,514 GitHub stars across the 10,000-plus tracked open-source AI projects, and 51 new projects entered the index. This week 2,592 new projects joined the index, plus 100 new Hugging Face spaces.

Releases of the day

Pydantic AI v2.32.0 shipped with features including suggesting known model names for invalid identifiers and support for xAI attachment search. LMDeploy v0.16.0 added support for Interns2 mobius and GLM-5.2. Cline desktop-v0.0.14 now posts native macOS notifications when a task finishes or needs input. Activepieces 0.88.2 widened action classification to READ, SEARCH, WRITE and DESTRUCTIVE categories. Langflow v1.11.4 fixed index alignment in release workflows and PyPI polling. LoopX v0.4.9 introduced PyPI-first installation and native Windows and KunlunCode hosts. SurfSense v0.0.37 updated documentation and sorted imports. n8n 2.35.4 fixed a bug in the API for decorator route schemas.

n8n [email protected] 2.35.4 (2026-08-19) · Bug Fixes · * API: Return a request schema for decorator routes in /discover (#36497) (fa781db) * Google Ads Node: Select the renamed video views metric... Langflow v1.11.4 What's Changed 🐛 Bug Fixes * fix: keep LFX index aligned in release workflows by @erichare in https://github.com/langflow-ai/langflow/pull/14105 * fix(ci): poll PyPI before... Cline desktop-v0.0.14 The app now posts native macOS notifications when a task finishes or needs your input, so you can leave Cline working in the background. Configure them under Settings →... Activepieces 0.88.2 * feat(framework): widen action classification to READ/SEARCH/WRITE/DESTRUCTIVE (#14852) @ibrahim-abuznaid, @claude * feat: remove s3 path for piece bundle (#14862) @MrChaker *... Pydantic AI v2.32.0 What's Changed 🚀 Features * Suggest known model names for invalid identifiers by @dsfaccini in https://github.com/pydantic/pydantic-ai/pull/7325 * Support xAI attachment search... SurfSense v0.0.37 What's Changed * docs: point CONTRIBUTING at the tooling the repo actually uses by @Yigtwxx in https://github.com/MODSetter/SurfSense/pull/1665 * style(web): sort imports in... LMDeploy v0.16.0 What's Changed 🚀 Features * Support Interns2 mobius by @RunningLeon in https://github.com/InternLM/lmdeploy/pull/4816 * feat: support GLM-5.2 by @CUHKSZzxy in... LoopX v0.4.9 LoopX v0.4.9 · At a Glance · PyPI-first installation, native Windows and KunlunCode hosts, a repository change-window that cannot silently lose gated code, and substantially...

Pricing moves

Seven price increases hit input costs today. GLM 5.2 from Z.AI saw the largest jump at 159 percent to $1.19 per 1M tokens. MiniMax M3 batch input rose 100 percent to $0.3 per 1M tokens. Kimi K2.7 Code batch from Moonshot AI rose 98 percent to $0.95 per 1M tokens. Nemotron 3 Ultra batch from NVIDIA rose 100 percent to $0.6 per 1M tokens. GLM 5.2 batch from Z.AI rose 100 percent to $1.4 per 1M tokens. Inkling batch from Thinkingmachines rose 100 percent to $1 per 1M tokens. DeepSeek V4 Flash from DeepSeek rose 14 percent to $0.08 per 1M tokens.

▲ 14%DeepSeek V4 Flash 0423DeepSeek · input $0.07 → $0.08 per 1M tokens
▲ 159%GLM 5.2Z.AI · input $0.46 → $1.19 per 1M tokens
▲ 100%MiniMax M3 (batch)MiniMax · input $0.15 → $0.3 per 1M tokens
▲ 98%Kimi K2.7 Code (batch)Moonshot AI · input $0.48 → $0.95 per 1M tokens
▲ 100%Nemotron 3 Ultra (batch)NVIDIA · input $0.3 → $0.6 per 1M tokens
▲ 100%GLM 5.2 (batch)Z.AI · input $0.7 → $1.4 per 1M tokens
▲ 100%Inkling (batch)Thinkingmachines · input $0.5 → $1 per 1M tokens

Research pick

A new paper called ASI-Bench proposes a benchmark for measuring progress toward artificial superintelligence. The authors argue that ASI requires AI to move beyond mastering existing knowledge toward exploring the unknown, creating new knowledge, and turning ideas into verifiable results. The paper notes that today's AI systems still fall short of that goal.

ASI-Bench: At the Dawn of Artificial SuperintelligenceJunwei Zhou, Zhen Sun, Binyu Li et al. · 38 community upvotes · Artificial superintelligence (ASI) requires AI to move beyond mastering existing knowledge toward exploring the unknown, creating new knowledge, and turning new ideas into verifiable results. However, the capabilities of today's AI systems

The bigger picture

The broad price increases across multiple providers and models suggest a market-wide adjustment in input costs, while the steady inflow of new projects and stars indicates continued expansion of the open-source AI ecosystem.

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