DeepSeek V4.1 Flash Is Now Free on OpenCode

OpenCode has made DeepSeek V4.1 Flash free to use. The free version supports a 200K context window. Its credit in the OpenCode Go plan was raised to $80 before, and calls no longer count against it.

Released in September, DeepSeek V4.1 Flash is a native multimodal MoE model with 552B main parameters and 196B extra memory parameters. Its causal encoder-decoder design cuts prefill compute nearly in half. DeepSeek's own platform still has no free API tier, and billing stays prepaid.

Speed is a hot topic among users. Some saw peaks of about 490 tokens per second on the ollama cloud platform, but speeds often dropped to 50-60 in the afternoon. Speed also falls as context grows, hitting about 50 tokens per second at 200K context.