Granite 4.2 30B

Granite 4.2 30B is a text chat model released by IBM as part of its Granite series of general-purpose language models. The name places it in the 4.2 generation of Granite at approximately a 30-billion-parameter scale, sitting in the mid-size range of the family — larger than the compact Granite variants intended for edge and CPU deployment, and smaller than the largest configurations IBM has shipped under the Granite name. IBM has historically developed Granite for enterprise and business-application workloads rather than as a consumer chat assistant. On measured serving performance, third-party benchmarking from Artificial Analysis records Granite 4.2 30B at roughly 77 output tokens per second with a time to first token of about 249 milliseconds. That combination — sub-quarter-second first-token latency and a throughput figure in the high-70s — puts it in a range suitable for interactive chat and streaming interfaces, though actual numbers will vary by hosting provider, hardware, batch size, and prompt length. We do not currently track a confirmed context window, modality set, or tool-calling specification for this entry, so buyers evaluating it for a specific integration should confirm those details against the serving provider's own documentation. In practice, a 30B-class chat model like Granite 4.2 30B is typically deployed where a mid-size model offers a better cost and latency profile than a much larger frontier model, while still handling multi-turn conversation, summarization, and instruction-following work. Compared with the very small Granite variants, it trades higher per-token serving cost for more headroom on reasoning-heavy prompts; compared with much larger models, it trades some capability ceiling for faster responses and cheaper inference. Check the pricing table on this page to see which providers currently serve it and how their rates compare.