DeepSeek V4 Flash input price fell 25% to $0.03 per 1M tokens, while GLM 5.3 from Z.AI jumped 536% to $1.4 and Kimi K3 from Moonshot AI surged 543% to $2.7. The ecosystem added 80,403 GitHub stars today across the 10,000+ tracked open-source AI projects, with 55 new projects entering the index. This week saw 2,837 new projects and 100 new Hugging Face spaces. OpenShell led GitHub with +3,902 stars.
AI News · October 3, 2026
DeepSeek V4 Flash drops 25%, GLM and Kimi prices soar
80,403 stars today55 new projects+3,902 OpenShell
Releases of the day
Windmill v1.822.0 introduced a cancel-only token scope that can be optionally path-scoped, giving users finer control over job cancellation permissions. Pydantic AI v2.54.0 added compatibility for draft-7 list-form items in JsonSchemaTransformer, TestModel, and Mistral streamed output, and made TestModel accept boolean values. AG2 v1.1.2 fixed MCP model output schema derivation from serialization and upgraded the Tenki SDK while preserving timeout settings. OpenClaw v2026.9.8 shipped with 58 commits and 43 pull requests from 21 contributors.
Windmill v1.822.0 feature1.822.0 (2026-10-03) · Features · * add a cancel-only jobs:cancel token scope, optionally path-scoped (#11479) (4bc7e0d) * add an instance setting routing all dependency jobs...
OpenClaw v2026.9.8 OpenClaw v2026.9.8 · 58 commits · 43 pull requests · 21 contributors · Changes included in this release · The release notes and changelog contain the same content, presented in...
Pydantic AI v2.54.0 What's Changed ⚠️ Compatibility Notes * Handle draft-7 list-form items in JsonSchemaTransformer, TestModel and Mistral streamed output, and make TestModel accept boolean...
AG2 v1.1.2 What's Changed * fix(mcp): derive model output schemas from serialization by @asts-top in https://github.com/ag2ai/ag2/pull/3317 * fix(tenki): upgrade sdk and preserve timeout...
Pricing moves
Input prices moved in opposite directions today. DeepSeek V4 Flash dropped 25% to $0.03, Gemma 4 26B A4B fell 12% to $0.07, Qwen3.5-35B-A3B decreased 6% to $0.15, Nemotron 3 Ultra declined 17% to $0.5, and Inkling fell 5% to $0.95. On the rise, GLM 5.2 increased 193% to $0.41, GLM 5.3 jumped 536% to $1.4, and Kimi K3 soared 543% to $2.7. Z.AI and Moonshot AI raised rates sharply while several other providers cut.
▼ 25%DeepSeek V4 Flash 0423DeepSeek · input $0.04 → $0.03 per 1M tokens
▲ 193%GLM 5.2Z.AI · input $0.14 → $0.41 per 1M tokens
▼ 12%Gemma 4 26B A4BGoogle · input $0.08 → $0.07 per 1M tokens
▼ 6%Qwen3.5-35B-A3BAlibaba · input $0.16 → $0.15 per 1M tokens
▼ 17%Nemotron 3 UltraNVIDIA · input $0.6 → $0.5 per 1M tokens
▲ 543%Kimi K3Moonshot AI · input $0.42 → $2.7 per 1M tokens
▼ 5%InklingThinkingmachines · input $1 → $0.95 per 1M tokens
▲ 536%GLM 5.3Z.AI · input $0.22 → $1.4 per 1M tokens
Research pick
OneStreamer, a paper by Xiangyu Zeng, Yuandong Yang, Zhiqiu Zhang and others, tackles streaming video LLMs that must retain evidence before its relevance is known and respond when enough evidence accumulates. The work proposes unifying perception, memory, and proactive response to form reusable factual memory without sacrificing real-time performance. The paper received 152 upvotes on Hugging Face.
The bigger picture
Today's price swings show a widening gap between aggressive cuts from major providers and steep hikes from others, while ecosystem growth remains steady. Research on streaming video interaction points to a growing focus on memory and real-time reasoning in LLMs.
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