Open weights
Moonshot releases Kimi K2.5 with open weights (opens in a new tab)
One of the strongest open options for agentic workloads, self-hosted or routed cheaply.
huggingface.co ·
MCP gets interactive apps, OpenAI's Responses shape becomes an open spec, Kimi K2.5 brings trillion-scale open weights, and Claude Code starts the 2.1 line.
One of the strongest open options for agentic workloads, self-hosted or routed cheaply.
Buildable into your own harness instead of reachable only through the Codex product.
blog.modelcontextprotocol.io
MCP Apps
One UI runs across hosts, so an MCP tool can ship its own interface.
A big release for skill and agent authors.
Matters if you maintain provider-abstraction layers: a candidate successor to cloning the Chat Completions schema.
That curve decides which agent workloads are affordable next year.
Useful context for reading between the lines of current MoE release notes.
Documented training and eval detail rather than a bare weights drop, if you run local or edge models.
Practical reading if you care about speed-versus-quality trade-offs in open models.
It sets the frame for the safety debates that will reach builders as policy.