[KongchangAI]
ModelMiMo

MiMo 2.6 Pro

A series of open-weight language models introduced by Xiaomi, focused on programming and reasoning capabilities. It reports relatively high scores on terminal benchmarks.

Timeline (last 90 days)

Nov 13

When independent evaluators run Terminal Bench 2.1, models such as MiMo 2.6 Pro and Qwen3 32B Max score in the 60s, far below the vendors' published values, and the evaluation framework itself is also changing, so top scores are not fully comparable.

Unverified50%
Sep 23

MiMo 2.6 Pro scored 46 points on the SWE-bench series of benchmark leaderboards, reaching the top of the global open-weight model rankings.

Expired50%
Sep 23

In a cycling animation test, MiMo 2.6 Flash used inverse kinematics to link pedal rotation with the joints of both legs, while Pro additionally incorporated richer lighting and shading layers along with frame details.

Unverified50%
Sep 23

In a test of folding a cross-shaped piece of paper into a cube, MiMo 2.6 Pro precisely derived the rotation axes and quaternion transformations, with the joints fitting together seamlessly.

Unverified50%
Sep 23

In a Blender floating island generation test, MiMo 2.6 Pro was significantly better than Flash in terms of scene completeness.

Unverified50%
Sep 23

Both MiMo 2.6 Flash and Pro successfully completed industrial-grade Agent tasks such as file notification delivery, reconciliation, and paginated order email processing.

Unverified50%

All Facts (5)

Source Articles