Developer
Xiaomi
8 indexed models; 7 currently meet the evidence threshold for the overall ranking. Together they have results from 16 evaluations.
Models by Xiaomi
| Model | Evals | Components | Rank | Release date |
|---|---|---|---|---|
| MiMo v2 Omni | 3 | 3/7 | 28 | 2026-03-18 |
| MiMo-V2.6-Pro | 7 | 7/7 | 65 | 2026-09-21 |
| MiMo v2 Pro | 5 | 5/7 | 67 | 2026-03-18 |
| MiMo v2.5 Pro | 9 | 6/7 | 68 | 2026-04-27 |
| MiMo v2.5 | 8 | 5/7 | 110 | 2026-04-27 |
| MiMo v2 Flash | 7 | 6/7 | 240 | 2025-12-16 |
| MiMo-V2.6-Flash | 6 | 7/7 | 250 | 2026-09-21 |
| MiMo-V2.6-Pro-UltraSpeed | 1 | 3/7 | — | 2026-09-21 |
Evaluations covering Xiaomi models (16)
AA-Omniscience · Adversarial Poetry Refusal (AHB self-run) · Arena Factuality — Text Arena (factuality-only weighting) · BullshitBench v2 · Concordia AI Risk Monitor · DystopiaBench · Enkrypt AI Safety Leaderboard · Inkling-Small model card — FORTRESS · Inkling-Small model card — StrongREJECT · LiveSecBench · Manager Coercion Bench · Opposite-Narrator Sycophancy · SM-Bench · SpeciEval · TAC · WildClawBench Safety & Alignment (OpenClaw harness)