Technical profile
A serving contract built around the model.
Kernelwright applies a repeatable qualification discipline across model launches: validate the model’s supported inputs, limits, streaming, tools, structured output, usage accounting, and failure behavior as one production interface. Performance statements follow end-to-end measurement.
Interface
Compatibility contract
The serving interface is designed for marketplace routing and direct product integration, with model-specific capabilities declared explicitly.
| Capability | Commitment | Standard |
|---|---|---|
| Chat completions | OpenAI-compatible request and response shapes | Core contract |
| Streaming | Server-sent events with measured time to first token | Core contract |
| Tool calling | Compatible tool definitions and structured calls | Model-qualified |
| Structured output | Schema-constrained response support | Model-qualified |
| Usage | Prompt, completion, and total-token accounting | Core contract |
| Admission control | Early 429 responses instead of unbounded queues | Core contract |
Qualification standard
What Kernelwright verifies
Prompt, completion, and total-token accounting
Time to first token and complete-response latency
Correctness across text, streaming, tools, and structured output
Request-size, model-limit, and media-input enforcement
Rate limits, error classes, terminal events, and recovery behavior
Public model metadata, pricing declarations, and capability accuracy
Capacity policy
No benchmark theater.
Published capacity will reflect sustained validated throughput with normal compatibility features enabled. If demand exceeds that capacity, the endpoint will reject early rather than hide latency in a long queue.