From the source
LG AI Research opened free API access to K-EXAONE 2.5 on November 9, 2026, letting developers, startups, companies and research institutions obtain an API key for service development and research through the EXAONE API Platform.
- LG describes K-EXAONE 2.5 as a technology-validation model produced by applying additional training to the version released in August 2026, and says it measured the model internally on open datasets and found it competitive with frontier-class global open-weight models.
- On LG's reported figures, K-EXAONE 2.5 improves on its predecessor across software development and modification, web information finding, SaaS tool automation and agent problem-solving; it scores above
DeepSeek-V4.1-Flash on SWE-Bench Pro, BrowseComp and Claw-Eval but below it on AutomationBench, and above GLM-5.3 only on AutomationBench.
- The model is text-only and supports Korean and English, with documentation listing 31 billion active parameters, a January 2026 knowledge cutoff, and chat, tool calling, structured outputs, code generation and streaming.
- Its endpoint reference specifies an
OpenAI-compatible POST /v1/chat/completions interface using the model identifier k-exaone_v2.5, where input tokens plus the max_tokens generation allowance cannot exceed 524,288 tokens and larger requests return a 400 error.
- Reasoning controls are exposed directly: reasoning_effort defaults to max, and preserve_thinking can carry earlier assistant reasoning into multi-turn requests, though LG's agent guide says higher reasoning intensity can increase response time and token usage.
- The guide includes configurations for the Hermes general agent and OpenCode coding agent, in which the external harness executes tool calls returned by the model and passes their results back.