Vani has added Kimi K3.1 Pro Soft-Derisked, a version of Moonshot AI's Kimi K3.1 Pro running with Vani's own soft-derisked instructions. It refuses less often than standard Kimi K3.1 Pro but is more conservative than GLM 5.3 Derisked. It's available in Vani Chat on Plus and higher, and via the API as moonshotai/kimi-k3.1-pro-soft-derisked, and it is for adults, with Vani asking users to confirm they are 18 or older. It has 1 million tokens of context, up to 131,072 output tokens, and always thinks first with low, high or max effort. API pricing is $5 per 1M input tokens, $0.50 cached, and $7 output, with Vani's instructions adding about 1,100 input tokens usually billed at the cached rate.
Questions this post answers
What is Kimi K3.1 Pro Soft-Derisked on Vani and how does it differ from standard Kimi K3.1 Pro?
It is the same Kimi K3.1 Pro model from Moonshot AI running with Vani's own soft-derisked instructions, which make it refuse less often than standard Kimi K3.1 Pro while staying more conservative than GLM 5.3 Derisked. It keeps 1 million tokens of context and up to 131,072 output tokens, and requires users to confirm they are 18 or older in Vani Chat. Developers evaluating refusal behavior across model variants can compare options like this through daily.dev.
What is the API pricing for moonshotai/kimi-k3.1-pro-soft-derisked on Vani?
Input tokens cost $5.00 per 1 million, cached input costs $0.50 per 1 million, and output costs $7.00 per 1 million. Vani's soft-derisked instructions add about 1,100 input tokens to each request, and these are usually billed at the cheaper cached rate rather than the full input rate. Teams budgeting API spend across model variants can track pricing changes like this via daily.dev.