Provider-documented facts. “Not collected” means we have not verified that field yet, not that the feature is unsupported.
Context / input limit
Not collected
No reviewed model-specific inference contract. Unknown does not mean unsupported.
Maximum input tokens
Not collected
No reviewed model-specific inference contract. Unknown does not mean unsupported.
Maximum output
Not collected
No reviewed model-specific inference contract. Unknown does not mean unsupported.
Parameters
Not collected
No reviewed model-specific inference contract. Unknown does not mean unsupported.
Release date
Not collected
No reviewed model-specific inference contract. Unknown does not mean unsupported.
Knowledge cutoff
Not collected
No reviewed model-specific inference contract. Unknown does not mean unsupported.
Tool calling
Not collected
No reviewed model-specific inference contract. Unknown does not mean unsupported.
Structured output
Not collected
No reviewed model-specific inference contract. Unknown does not mean unsupported.
Streaming
Not collected
No reviewed model-specific inference contract. Unknown does not mean unsupported.
Batch processing
Not collected
No reviewed model-specific inference contract. Unknown does not mean unsupported.
Fine-tuning
Not collected
No reviewed model-specific inference contract. Unknown does not mean unsupported.
Reasoning
Not collected
No reviewed model-specific inference contract. Unknown does not mean unsupported.
Context caching
Not collected
No reviewed model-specific inference contract. Unknown does not mean unsupported.
Base model
Not collected
The source compares the RFT model against base o3-mini, but does not explicitly identify the exact trained ancestor or snapshot. A comparison baseline is not conclusive lineage evidence.
Historical private RFT evaluation. Current access to this exact trained variant is unverified; historical does not assert retirement.
Regions
Model-specific processing and deployment regions require confirmation.
Endpoint pricing has not been collected. Check the provider for applicable rates.
Data & enterprise terms
Runloop public terms retain service IP and customer ownership of submitted data. The privacy policy uses purpose-based retention and allows US and other-country processing. These documents do not establish a model-specific no-training guarantee, zero retention or cache lifetime.
Caching can reuse earlier input processing. Availability, lifetime and billing depend on the endpoint. A cache hit still generates a new response.
01 / INITIAL INPUTProcess & retain
The system processes eligible input and retains reusable state.
02 / LATER REQUESTCheck available state
Reuse depends on matching input, session state and the cache policy.
03 / CACHE HITReuse & generate
Reusing state can reduce repeated work. Any charges follow the billing rules below.
Runloop
Not verified
No reviewed model inference-cache contract. Shared design context and reusable sandbox blueprints are not proof of prompt caching.
How it works
Not collected
No reviewed model inference-cache contract. Shared design context and reusable sandbox blueprints are not proof of prompt caching.
Minimum reusable input
Not collected
No reviewed model inference-cache contract. Shared design context and reusable sandbox blueprints are not proof of prompt caching.
Lifetime & refresh
Not collected
No reviewed model inference-cache contract. Shared design context and reusable sandbox blueprints are not proof of prompt caching.
How to enable it
Not collected
No reviewed model inference-cache contract. Shared design context and reusable sandbox blueprints are not proof of prompt caching.
Usage fields
Not collected
No reviewed model inference-cache contract. Shared design context and reusable sandbox blueprints are not proof of prompt caching.
Billing rules
Not collected
No reviewed model inference-cache contract. Shared design context and reusable sandbox blueprints are not proof of prompt caching.
Limits & data boundaries
Not collected
No reviewed model inference-cache contract. Shared design context and reusable sandbox blueprints are not proof of prompt caching.
No verified cache rate for this endpoint. An empty rate is not a zero charge.
Caching reviewed 16 Oct 2026
What the measurements say
No benchmark observations have been added for this model. We do not turn a vendor description into a performance score.
Latency and throughput depend on the host, request size, reasoning settings and load. There is no measured speed or global usage figure in this record.
The record behind the model
Why it is included
Official OpenAI case study documents a privately trained reinforcement-fine-tuned model. No public derivative weights or exact standalone checkpoint selector established.
