RightNow AI launches GPU-optimized inference for open models
Start here
RightNow AI has launched an inference service for open models, tuning decoding, precision and engine settings per model and hardware, and publishing per-million-token prices alongside measured speeds. Its table lists DeepSeek V4.1 Flash at USD 0.15 input and USD 0.60 output per million tokens at 327 tokens per second, and Nemotron 3.5 Lightning 30B at USD 0.05 input and USD 0.15 output at 540 tokens per second. Billing is pay-as-you-go per token with no subscription, on shared capacity; reserved capacity and dedicated single-company deployments are listed as not yet offered. Prompt caching is automatic and best effort, requiring no header or parameter, the API accepts Chat Completions and Messages requests, and new accounts receive $1 free.
Generated from reports · updated 2 h ago
Report timeline
Follow the reports to see the event from different sides.
- Hacker News · AIRightNow AI launches GPU-kernel-optimized inference service for open models
RightNow AI launched an inference service that runs open models with decoding, precision and engine settings tuned per model and hardware, and it published per-million-token prices alongside measured speeds. Its table lists DeepSeek V4.1 Flash at USD 0.15 input and USD 0.60 output per million tokens at 327 tokens per second, and Nemotron 3.5 Lightning 30B at USD 0.05 input and USD 0.15 output at 540 tokens per second. Prompt caching is automatic and best effort with no header or parameter required, the API accepts Chat Completions and Messages requests, and new accounts get $1 free.
Heat for this event
Not enough continuous observations yet to draw a trend.