Open-source AI
The open-source side of the model release cycle: open-weights models, the engines and tooling to run them, and the agent ecosystem built around them.
Last updated: Nov 18, 2026
What belongs in this section
This is where Modelspike covers the open-source world around AI models:
- Open-weights models — any release with downloadable weights, recorded with its license and hardware requirements.
- Inference and deployment engines — runtimes, quantization and serving tooling for open models.
- Agent and coding ecosystem — agent frameworks, coding agents and plugin ecosystems. These live on inner pages only, never on the homepage.
What is out of scope
Non-AI open-source projects — frontend frameworks, self-hosted apps and the like — are not covered on this site. The tracker stays on AI models and the ecosystem directly around them.
What is listed so far
Both models currently tracked — GPT-6.2 Astra and GPT-6.2 Sol — are proprietary, so the open-weights list is empty today. It fills in as open-weights releases land: each entry gets the same treatment as every model page, plus a weights link, license, hardware estimate and a minimal run example in the how-to-use section.
FAQ
Do open-weights models get the same treatment as proprietary ones?
Yes — open-weights releases get the same four modules (benchmarks, comparison, how-to-use, API pricing), plus a weights link, license, hardware estimate and quantization notes in the how-to-use section.
Are coding agents and CLI tools covered here?
The agent and coding ecosystem is covered on inner pages in this section — but never on the homepage, which stays focused on model releases.