On October 15, OpenCode announced that DeepSeek V4.1 Flash is now free, with the free tier supporting a 200K context window, lower than the 1M-context version. After the free rollout, calls to the model on OpenCode no longer count toward quota, extending the earlier Operation Cheepseek plan.
MLCommons released the MLPerf Inference v6.1 benchmark suite. AMD entered with 512 Instinct MI355X GPUs, scoring 2,901,950 tokens/s offline and 2,405,310 tokens/s in server mode on DeepSeek R1, setting a record for AMD's largest submission. NVIDIA submitted 288-GPU configurations of GB300 and GB200, and its Vera Rubin VR200 appeared in the benchmark for the first time.