OPEN LINK ↗
// 56m agoBENCHMARK RESULT
Celeris-1 Decision Tops Speed-Intelligence Frontier
Artificial Analysis places Celeris-1 Decision on the speed-intelligence Pareto frontier, measuring it at 0.06 seconds per task—more than twice as fast as Jev 1.13. Its Decision Index score also sits four points above GPT-6 Luna, showing that low-latency structured inference can compete with general-purpose models on targeted workflows.
// ANALYSIS
This is a strong case for treating decision models as a distinct production primitive, especially where latency and predictable outputs matter more than open-ended generation.
- –Fixed choices and probabilities let developers threshold, rank, and route results without parsing free-form text.
- –Celeris supports text, structured JSON, and images, with optional explanations for human review and auditing.
- –Its hosted API costs $0.04 per million input tokens, with output priced free.
- –The benchmark favors classification and tool-calling workflows; Celeris is not a universal replacement for stronger reasoning models.
- –The clearest use cases are agent control, support routing, fraud checks, and real-time voice systems.
// TAGS
celeris-1-decisionbenchmarkevaluationinferencestructured-outputtool-usellm
DISCOVERED
56m ago
2026-10-15
PUBLISHED
1h ago
2026-10-14
RELEVANCE
9/ 10
AUTHOR
tom_w_hamer