LG AI Research / lgai/k-exaone-2.5

K-EXAONE 2.5 - access through LLMTR

K-EXAONE 2.5 is LG AI Research's 750-billion-parameter text model with 31 billion active parameters. It was trained for Korean and English and also answers Turkish prompts; its knowledge cutoff is January 2026. Its context window is 524,288 tokens, and the prompt plus max_tokens cannot exceed it. It accepts text only. Tool calling, JSON mode and structured output with json_schema are supported. It reasons before every answer and reasoning cannot be turned off: choose low, high or max with reasoning_effort; the default is max. The reasoning text arrives in reasoning_content and reasoning tokens count toward max_tokens. In multi-turn agent flows, send chat_template_kwargs.preserve_thinking: true to keep the reasoning of earlier turns. It is free with a daily allowance that counts requests and input and output tokens, reasoning included. When the allowance runs out, wait for the time given in the response and try again.

Technical specifications

Canonical IDlgai/k-exaone-2.5
ProviderLG AI Research
Context window524,288 tokens
OperationsCHAT_COMPLETIONS
Modalitiestext

Pricing

An 8% platform margin applies to credit top-ups; model usage prices are not separately marked up.

OperationMetricUnitPrice
CHAT_COMPLETIONSINPUT_TEXTPER_1M_TOKENSNot available
CHAT_COMPLETIONSOUTPUT_TEXTPER_1M_TOKENSNot available

Example usage

With existing OpenAI SDK flows, change only the base URL and model identifier.

curl https://llmtr.com/v1/chat/completions   -H "Authorization: Bearer llmtr-your_key"   -H "Content-Type: application/json"   -d '{"model":"lgai/k-exaone-2.5","messages":[{"role":"user","content":"Hello"}]}'

Related models