LG AI Research / LG internal CodeLlama adaptations
Overlapping-context adaptation evaluated in LG’s ACL 2025 industry paper.
Provider-documented facts. “Not collected” means we have not verified that field yet, not that the feature is unsupported.
Internal deployment, infrastructure costs and published efficiency improvements are not a per-token offer.
Conference publication month; checkpoint creation date unknown.
Training setup, not a verified deployment context limit.
Prices belong to an endpoint and billing tier. A provider listing does not guarantee access for every account or region.
Not collectedNo public request selector verified.
Endpoint pricing has not been collected. Check the provider for applicable rates.
Retention, training use, regional processing and enterprise terms have not been summarized for this endpoint.
Caching can reuse earlier input processing. Availability, lifetime and billing depend on the endpoint. A cache hit still generates a new response.
The system processes eligible input and retains reusable state.
Reuse depends on matching input, session state and the cache policy.
Reusing state can reduce repeated work. Any charges follow the billing rules below.
No verified cache rate for this endpoint. An empty rate is not a zero charge.
Table 4; stride 512; best checkpoint, epoch 1.2. Private 100-function coding benchmark; not comparable to public HumanEval.
Evaluation source (opens in a new tab)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.
Internal trained configuration documented by LG authors. Exact adapted weights are absent from the reviewed official release inventories. Classification is inferred for this configuration; it does not classify the public CodeLlama base as closed.
Classification evidence (opens in a new tab)