RLTasksby ulam.ai
Models / Xiaomi

Xiaomi / MODEL EVIDENCE

MiMo V2.6 Pro

60 benchmarks · 80 observations · 55 sources · released 2026-09-21

Scores at a glance

52 benchmarks · large dot: MiMo V2.6 Pro; small dots: closest models where reported

AA-3D ↗983.14 Elo
572.551198.83

MiMo-V2.6-Pro · AA-3D · Elo ↗

AA-PersonalityBench ↗28.92 percentile
28.92100

MiMo-V2.6-Pro · AA-PersonalityBench · expressiveness ↗

7 more reported results

28.92 percentile · MiMo-V2.6-Pro · AA-PersonalityBench · organization · Artificial Analysis ↗

16.23 percentile · MiMo-V2.6-Pro · AA-PersonalityBench · informativeness · Artificial Analysis ↗

44.15 percentile · MiMo-V2.6-Pro · AA-PersonalityBench · transparency · Artificial Analysis ↗

54.31 percentile · MiMo-V2.6-Pro · AA-PersonalityBench · warmth · Artificial Analysis ↗

16.23 percentile · MiMo-V2.6-Pro · AA-PersonalityBench · conviction · Artificial Analysis ↗

69.54 percentile · MiMo-V2.6-Pro · AA-PersonalityBench · imagination · Artificial Analysis ↗

59.38 percentile · MiMo-V2.6-Pro · AA-PersonalityBench · slop · Artificial Analysis ↗

STEMBench ↗2.172 / 4
2.172.68

MiMo V2.6 Pro · STEMBench120 · mean progress / 4 · closed ↗

ErdősBench ↗3.020 / 4
2.113.25

MiMo v2.6 Pro · ErdősBench · mean progress / 4 · API ↗

SAFE-Teen ↗40%
1556

MiMo v2.6 Pro · Urgent safety · share of applicable ratings marked fail ↗

2 more reported results

57% · MiMo v2.6 Pro · Health advice · share of applicable ratings marked fail · Vals AI · SAFE-Teen ↗

56% · MiMo v2.6 Pro · Relationship boundaries · share of applicable ratings marked fail · Vals AI · SAFE-Teen ↗

ALE-Bench ↗1158.1 Performance points
137.782951.32

mimo-v2.6-pro · Average contest performance ↗

Closest models by score

Similar positions on shared benchmark leaderboards. Smaller gaps are closer.

How similarity is measured

Mean absolute gap in percentile positions on shared benchmark settings; lower is closer. At least 60% overlap is required, with a minimum of three when available. Independent evaluations are used when there are at least three; otherwise published report comparisons are used. Repeated model observations within one setting are averaged before ranking. Writing-style traits and composite indices are excluded. This describes similarity, not an overall quality ranking.

POST-TRAINING / SOURCE DISCLOSURE

What the MiMo-V2.6 Pro report says

Xiaomi describes large-scale multitask reinforcement learning with asynchronous generation, varied agent harnesses and group-based grading. Freezing the MoE router is one of the stated stability measures.

  • Multitask reinforcement learning
  • Asynchronous rollout infrastructure
  • Multiple agent harnesses and group-based grading
  • Frozen MoE router during the reported scale-up
Stateful agent training →Reward design · background →

The reviewed announcement describes the scale-up and stability mechanisms. Frozen routing is not the same as routing replay, and no specific policy loss is inferred from the presence of group grading.

  1. MiMo-V2.6: Scaling Up Reinforcement Learning for Self-Improvement ↗

    Xiaomi MiMo Team · 2026

Filter comparisons to this model →

Sources

Every plotted observation links to its evaluation. Reported configurations, benchmark versions and task subsets can differ; similarity describes shared scores, not training lift.

  1. Arena leaderboard — WebDev evaluation ↗

    Arena

  2. AA-3D: Interactive 3D Experience Benchmark | Artificial Analysis ↗

    Artificial Analysis

  3. AA-AnalystAgent Benchmark | Artificial Analysis ↗

    Artificial Analysis

  4. AA-Briefcase v1.1: Agentic Knowledge Work Benchmark | Artificial Analysis ↗

    Artificial Analysis

  5. AA-PersonalityBench: AI Personality & Communication Style Benchmark | Artificial Analysis ↗

    Artificial Analysis

  6. Artificial Analysis Intelligence Index v5.0 — Artificial Analysis ↗

    Artificial Analysis

  7. AA-LCR v1.1: Long Context Reasoning Benchmark | Artificial Analysis ↗

    Artificial Analysis

  8. AutomationBench-AA: Agentic SaaS Workflow Benchmark | Artificial Analysis ↗

    Artificial Analysis

  9. CritPt Benchmark | Artificial Analysis ↗

    Artificial Analysis

  10. CritPt v2 — Hard Subset — Artificial Analysis ↗

    Artificial Analysis

  11. FrontierCode 1.1 Benchmark | Artificial Analysis ↗

    Artificial Analysis

  12. FrontierCode 2.0 Benchmark — Artificial Analysis ↗

    Artificial Analysis

  13. GDP.pdf Benchmark | Artificial Analysis ↗

    Artificial Analysis

  14. GDPval-AA v2.1 Benchmark | Artificial Analysis ↗

    Artificial Analysis

  15. HLE-Diamond Benchmark — Artificial Analysis ↗

    Artificial Analysis

  16. Humanity's Last Exam Benchmark | Artificial Analysis ↗

    Artificial Analysis

  17. MLCR-AA: Medical Long Context Reasoning Benchmark | Artificial Analysis ↗

    Artificial Analysis

  18. AA-Omniscience v1.1: Knowledge and Hallucination Benchmark — Artificial Analysis ↗

    Artificial Analysis

  19. SciCode Benchmark | Artificial Analysis ↗

    Artificial Analysis

  20. Terminal-Bench-Science 0.1 Benchmark | Artificial Analysis ↗

    Artificial Analysis

  21. Terminal-Bench 4.0 Benchmark | Artificial Analysis ↗

    Artificial Analysis

  22. Ale Bench — methodology and published results ↗

    Epoch AI · 2026

  23. APEX-Agents ↗

    Epoch AI

  24. CritPt v1 ↗

    Epoch AI

  25. Humanity's Sixth Sense — evaluation methodology and results ↗

    Scale Labs / SEAL · 2026

  26. ErdősBench: research mathematics, scoped progress and proof audits ↗

    Ulam AI · 2026

  27. STEMBench — Six-Domain Model Evaluation | Ulam ↗

    Ulam

  28. Code Migration v1 — evaluation methodology and results ↗

    Vals AI · 2026

  29. Code Migration — evaluation methodology and results ↗

    Vals AI · 2026

  30. CyberBench — evaluation methodology and results ↗

    Vals AI · 2026

  31. Excel Modeling Benchmark v1 — evaluation methodology and results ↗

    Vals AI · 2026

  32. Excel Modeling Benchmark — evaluation methodology and results ↗

    Vals AI · 2026

  33. Finance Agent v2 — evaluation methodology and results ↗

    Vals AI · 2026

  34. Finance Agent — evaluation methodology and results ↗

    Vals AI · 2026

  35. Vals Finance Index — evaluation methodology and results ↗

    Vals AI · 2026

  36. HLAB-Vals-1.1 — evaluation methodology and results ↗

    Vals AI · 2026

  37. Harvey's Legal Agent Benchmark v1 — evaluation methodology and results ↗

    Vals AI · 2026

  38. IOI — evaluation methodology and results ↗

    Vals AI · 2026

  39. Vals Legal Index — evaluation methodology and results ↗

    Vals AI · 2026

  40. Legal Research Bench — evaluation methodology and results ↗

    Vals AI · 2026

  41. MedCode — evaluation methodology and results ↗

    Vals AI · 2026

  42. MedScribe — evaluation methodology and results ↗

    Vals AI · 2026

  43. SAFE-Teen: Teen Mental Health & Safety | Vals AI A conversation branches after the chatbot turns down a request — evaluation methodology and results ↗

    Vals AI · 2026

  44. MysteryMechanism — evaluation methodology and results ↗

    Vals AI · 2026

  45. ProgramBench — evaluation methodology and results ↗

    Vals AI · 2026

  46. ProofBench — evaluation methodology and results ↗

    Vals AI · 2026

  47. Public Benefits Bench — evaluation methodology and results ↗

    Vals AI · 2026

  48. SAGE — evaluation methodology and results ↗

    Vals AI · 2026

  49. SRE Bench — evaluation methodology and results ↗

    Vals AI · 2026

  50. SWE-bench Multimodal — evaluation methodology and results ↗

    Vals AI · 2026

  51. Tax Agent Bench — evaluation methodology and results ↗

    Vals AI · 2026

  52. Terminal-Bench 4.0 — evaluation methodology and results ↗

    Vals AI · 2026

  53. Terminal-Bench-Science — evaluation methodology and results ↗

    Vals AI · 2026

  54. Vals Index — evaluation methodology and results ↗

    Vals AI · 2026

  55. Vibe Code Bench — evaluation methodology and results ↗

    Vals AI · 2026

Names used in source tables

MiMo V2.6 Pro · MiMo V2.6 Pro (Mini-SWE-agent) · MiMo V2.6 Pro (OpenHands) · MiMo v2.6 Pro · MiMo-V2.6 Pro · MiMo-V2.6-Pro · MiMo-v2.6-Pro (high) · mimo-v2.6-pro