Understand this story
On October 12, 2026, Fireworks AI published a technical report on the Kimi K3 reinforcement learning infrastructure, describing training, rollout, and policy update optimizations for full-parameter reinforcement learning on the 2.8T-parameter Kimi K3. The report provides measured speedups for its in-house MoE Mega Kernel, KDA, and Attention Residuals kernels, and states that XOR delta compression reduces the average delta of a 1,024 GiB checkpoint to 20.3 GiB. The report also mentions that this infrastructure supported Harvey Tenet in achieving SOTA on LAB Contracts in the August 2026 evaluation, with a LAB all-pass rate of 19.7%, 1.8 times that of base K3. Currently, this report is the only disclosed progress, and there are no subsequent developments as facts.
AI-generated from public reports
Latest development · Report
Fireworks Releases Kimi K3 Large-Scale Reinforcement Learning Technical ReportDevelopments
1 confirmed developments
Fireworks publishes Kimi K3 reinforcement learning infrastructure report
Source:Fireworks AI
Reports and discussion timeline
Browse reports and discussion by time. Their evidence levels remain distinct.
Oct 12
Fireworks AIFirst-party
Fireworks Releases Kimi K3 Large-Scale Reinforcement Learning Technical Report
Fireworks released a technical report describing training, rollout, and policy update optimizations for full-parameter reinforcement learning on the 2.8T-parameter Kimi K3. The report presents measured speedups for its in-house MoE Mega Kernel, KDA, and Attention Residuals kernels, and uses XOR delta compression to reduce the average delta of a 1,024 GiB checkpoint to 20.3 GiB. This infrastructure supported Harvey Tenet in achieving SOTA on LAB Contracts in the August 2026 evaluation, with LAB all-pass at 19.7%, 1.8 times that of base K3.