Fetching from the wire…
01
02
03
04
05
06
07
08
Source-backed findings, relationship evidence, citations, and briefing history from the public MindPattern archive.
Composer 2 scores 61.3 on CursorBench vs Claude Opus 4.6's 58.2.
Source findingCursor released Composer 2, proprietary coding model trained via reinforcement learning.
Source findingComposer 2 is a fine-tuned variant of Moonshot's Kimi K2.5
Source findingCursor 3 ships with Composer 2 model based on fine-tuned Kimi K2.5
Source findingCursor Composer 2 uses compaction-in-loop reinforcement learning to reduce context errors.
Source findingCursor improves Composer 2 through reinforcement learning.
Source findingComposer 2 outperforms GPT-5.4 at 10x lower cost
Source findingComposer 2 beats Opus 4.6 on coding benchmarks at lower cost
Source findingCursor released Composer 2, proprietary coding model trained via reinforcement learning.
Source findingCursor 3 ships with Composer 2 model based on fine-tuned Kimi K2.5
Source findingCursor Composer 2 uses compaction-in-loop reinforcement learning to reduce context errors.
Source findingCursor improves Composer 2 through reinforcement learning.
Source finding