TECH Signal 494
Moonshot AI releases 2.8-trillion-parameter Kimi K3, the largest open-weight model, rivaling proprietary frontiers
↗
Moonshot AI released Kimi K3, a 2.8-trillion-parameter open-weight model that is the largest released as of August 2026, with 104B active parameters per token across 896 experts and a 1,048,576-token context window, benchmarking competitively against Claude Fable 5 and GPT-5.6 Sol.
Kimi K3 offers frontier-class performance at open-weight API pricing ($3/$15 per 1M tokens), but self-hosting requires at least 8 enterprise-grade accelerators with a hardware bill in the high six figures, pushing most teams toward API access. The custom 'Kimi K3 License' rather than plain MIT may also constrain some commercial adoption paths.
Written by elseif from the cluster below · every claim links back to a sourceThe three things worth knowing
Kimi K3 is the largest open-weight model at 2.8 trillion parameters, with 104B active per token across 896 experts and a 1,048,576-token context window.
API access costs $3/$15 per 1M tokens through most providers, while self-hosting requires 8+ enterprise accelerators at high six-figure cost.
Benchmarks place it alongside or behind Claude Fable 5 and GPT-5.6 Sol across coding, agent, and visual tasks, beating both on several specific tests.
THE CLUSTER