Fireworks makes Rune, a 26B decision model, live on serverless
Article timeline
The articles in this story. Times are ET.
Sophia YangXRune, a 26B-parameter System One decision model, launches on Fireworks AIAIRune is a System One decision model from Rune (@runedotgl) with 26B total parameters and 4B active per token, now live on Fireworks serverless. Per Fireworks AI, it reads a document once and answers up to 256 multiple-choice, true/false, or graded-scale questions in one pass, each with a calibrated probability. Pricing is $0.042 per 1M input tokens, $0.008 cached, and $0 output, with median server-side latency under 150ms and a 262K text-only context.
FireworksOfficialFireworks launches Rune, a 26B decision model for serverless useAIFireworks says its Rune model, a 26B-parameter System One decision model with 4B active per token, is live on serverless. It reads a document once and answers up to 256 multiple-choice, true/false, or graded questions, each with a calibrated probability. Median latency is under 150ms server-side, with a 262K text context and input priced at $0.042 per 1M tokens, $0.008 cached, and $0 for output.
