OPEN LINK ↗
// 46m agoINFRASTRUCTURE
DeepInfra Adds Kimi K3.1 Pro
DeepInfra now offers Day 0 access to Moonshot’s 2.8T-parameter MoE model with vision and a 1M-token context window, hosted on NVIDIA Blackwell in the US. The launch includes zero data retention and promotional pricing of $1.90 per million input tokens and $9.50 per million output tokens.
// ANALYSIS
DeepInfra turns an enormous open-weight model into a practical API option, making K3.1 Pro more accessible to developers who cannot self-host trillion-parameter infrastructure.
- –1M-token context and vision support target long-running coding, research, and agent workflows
- –US hosting and zero data retention are meaningful differentiators for privacy-sensitive teams
- –The model’s scale makes managed inference far more realistic than local deployment for most developers
- –Promotional pricing is attractive, but teams should budget against standard rates after the discount ends
// TAGS
kimi-k3.1-prollmopen-weightsmoelong-contextvisioninferencehosted-service
DISCOVERED
46m ago
2026-11-04
PUBLISHED
1h ago
2026-11-04
RELEVANCE
9/ 10
AUTHOR
DeepInfra