Morph v3 Large

Morph V3 Large is positioned as the high-accuracy tier of Morph's V3 lineup, built around an "apply" workflow that takes a natural-language instruction together with an existing code block and returns a precise transformation of that code. Gateway descriptions frame it as targeting complex code edits where depth and quality matter more than raw speed, distinguishing it from a lighter sibling, Morph V3 Fast. The model expects a structured prompt with an <instruction> section containing the edit directive and a <code> section holding the source snippet to be modified, a convention that signals it is designed for inline, tool-assisted code rewriting rather than open-ended chat or generation from scratch.

In the broader Morph family, V3 Large sits alongside V3 Fast as the more capable option for workloads where edit precision is critical, while Fast is oriented toward quicker iterations. Third-party listings describe throughput on the order of roughly 4,500 tokens per second with a reported 98% accuracy on targeted code transformations, figures that suggest the model is optimized for latency-sensitive developer tooling such as agentic refactors, automated patch application, and large-scale code migration tasks. Practically, it fits teams that need reliable, instruction-driven code modification at scale, accepting text prompts with embedded code and producing edited text output suitable for direct application to a codebase.