Model▪Risk▪Index

Rankings / Xiaomi

Xiaomi MiMo v2.5

Tier 025/100open weights

mimo-v2.5 (open weights, MIT) via mimo.mi.com / multiple hosts

Usage share 8.8% · OpenRouter rankings API (daily token share, 2026-08-04)

Second-largest tracked share. The under-review data-handling grade is honest, not evasive: the terms exist but resist automated capture, which makes this entry the first concrete requirement for the Phase 2 browser-rendering snapshot path. The closed sibling MiMo-V2.5-Pro (API-only, 1T params) is not covered by this entry.

Tier assessment

Fails one or more Tier 1 requirements: no published model card or safety evals, or terms that permit training on customer API data by default with no opt-out.

Tier 1 requirements
  • ✓Published model card. A model card or equivalent technical documentation is published for this model.
  • ✕Published safety evals. No safety evaluations are published for this model.
  • ✕Documented safety policy. No documented safety or acceptable-use policy exists.Terms pages exist but content is unverifiable by capture; MIT weights carry no acceptable-use policy.
  • ✕Enterprise data controls. Terms permit training on customer data by default with no documented opt-out.Hosted-API terms unverified; treated as unsatisfied pending Phase 2 browser capture.
Tier 2 requirements
  • ✕External pre-deployment testing. No disclosed external pre-deployment testing.
  • ✕Third-party certification. No verifiable third-party certification.
  • ✓Versioning with changelogs. Model versions are explicitly identified and changes are changelogged.Dated checkpoints and an update log — undermined in practice by documented silent updates under stable names.
  • ✕Stated deprecation policy. No stated deprecation policy.Deprecations are announced with dates, but no formal policy or notice window is published.

Missing for Tier 1: published safety evals, documented safety policy, enterprise data controls. The tier is computed from this checklist — satisfying these requirements moves the badge, automatically.

Risk analysisfive vectors · click a wedge for its evidence
DataStabilityAdversarialTransparencyCompliance

Risk vectors — the receipts

Data governanceunder review

What happens to your data: training-on-customer-data defaults, retention windows, residency options, and the enterprise-versus-consumer terms gap. A legal property, not a capability — it survives every model generation.

The hosted platform publishes privacy and user-agreement pages, but their JS-rendered content could not be captured for verification; secondary reporting claims a no-train API default with 30-day prompt logging, unconfirmed against Xiaomi's own text. Residency undisclosed. Self-hosting the MIT weights is deployer-controlled.

Receipts (1)

Operational stabilitypartial

Whether it changes without warning: versioning discipline, changelog quality, deprecation policy, and observed silent changes. The signal no one else tracks.

Genuinely mixed: dated checkpoints and a real update log with enforced dated deprecations (V2 series retired 2026-06-30) — but also documented silent in-place updates where 'model API call method and model name remain unchanged' while behavior shifted.

Receipts (1)

Adversarial resistanceweak

Whether an attacker can make it misbehave — direct jailbreaks against the model's own policies and indirect prompt injection in agentic tool use. Graded to the weaker of the two, because an attacker takes the easier path.

Jailbreak resistanceweak

First independent datapoint arrived in July 2026: CASI 73.80, bottom tier against a frontier band of 85–95 and near DeepSeek R1's historically poor score; Xiaomi publishes no safety evals and jailbreak system prompts circulate publicly.

Prompt injection (agentic)weak

Marketed on frontier-level agentic capability with 1,000+ tool-call runs in internal tests, yet no injection hardening or agent-security results have been published by anyone.

Receipts (3)

Transparencypartial

Whether you can see how it was built and tested: model cards, published safety evals, external pre-deployment testing, and disclosure of changes. The mechanism behind the tier ladder.

MIT open weights (310B MoE) and a maintained model-update log, but no technical report, no safety section, and no external pre-deployment testing.

Receipts (1)

Compliance postureweak

Whether it is certified and compliant: SOC 2, ISO/IEC 42001, HIPAA eligibility, EU AI Act readiness, and audit availability.

Xiaomi's corporate trust center lists company-level certifications, but nothing scopes the MiMo platform or API specifically.

Receipts (1)
  • Xiaomi trust center (corporate-level only)
    Xiaomi MiMo provider artifacts · provider artifact · source tier B · trust.mi.com · retrieved 2026-08-05
    “We fully demonstrate the compliance of our practices through regular self-assessment, third-party audits and certifications.”

Governance & evidence

Where your data goes

  • PRC

No regional pinning — the provider chooses where data is processed.

Residency undisclosed; the provider's PRC home jurisdiction is used here explicitly as a placeholder pending verification of the hosted platform's terms.

Enterprise vs consumer terms

Enterprise vs consumer gap: not assessed. Not yet assessed. Absence of assessment is not absence of a gap.NOT ASSESSEDENTERPRISE / APICONSUMER
Not assessed

Not yet assessed. Absence of assessment is not absence of a gap.

Change cadence

no tracked changesNo tracked changes for Xiaomi MiMo v2.5. Absence of detection is not evidence of stability.none

No tracked changes for this model. That is absence of detection, not evidence of stability — it may mean the model is unwatched, not that it is unchanging.

Score volatility

No dated score readings recorded for this model yet. Readings are only entered where multiple real, dated third-party values exist — never interpolated.

Receipts — what backs this assessment

8 evidence refs3 distinct sources1 independent
  • CF5 Labs CASI/ARS leaderboardindependent eval×1 reference
  • BXiaomi MiMo provider artifactsprovider artifact×6 references
  • EOpenRouter model rankingsusage data×1 reference

retrieved 2026-08-05

Compliance & deployment

Trains on customer data by default
not verified
SOC 2
not verified
ISO/IEC 42001
not verified
HIPAA eligible
not verified
Retention window
Unverified (secondary reporting claims 30-day prompt logs)
Data residency
Undisclosed
EU AI Act
No published EU AI Act posture.
Deprecation policy
Dated deprecations announced per-release; no formal policy
Available via
mimo.mi.com (hosted) · Self-hosted (MIT weights) · Multiple inference providers

Change timeline

No tracked changes yet for this model.

Compare this model: Xiaomi MiMo v2.5 + open compare view →