ModelRiskIndex

The wall

Every model, one plate each

One card per tracked model, generated from the registry rather than drawn. A dashed row is a vector we have not assessed — an absence, not a low score. The strip along the bottom edge is one mark per piece of evidence, gold where the claim is bound to a verbatim quote. The ornament is a guilloche seeded by those same evidence hashes: the figure cannot be produced without the provenance behind it, and no two models share one.

Cards fade as their evidence ages. Each is also served as an image — append .png to any card URL.

MODELRISKINDEX TIER 0 DeepSeek V4 Flash DEEPSEEK · deepseek-v4-flash via api.deepseek.com (alias silently r… DATA GOVERNANCE WEAK OPERATIONAL STABILITY WEAK ADVERSARIAL RESISTANCE WEAK TRANSPARENCY PARTIAL COMPLIANCE POSTURE WEAK 12/19 CLAIMS QUOTE-BOUND EVIDENCE 2026-08-05
MODELRISKINDEX TIER 0 Xiaomi MiMo v2.5 XIAOMI · mimo-v2.5 (open weights, MIT) via mimo.mi.com / multiple h… DATA GOVERNANCE PARTIAL OPERATIONAL STABILITY PARTIAL ADVERSARIAL RESISTANCE WEAK TRANSPARENCY PARTIAL COMPLIANCE POSTURE WEAK 14/19 CLAIMS QUOTE-BOUND EVIDENCE 2026-08-07
MODELRISKINDEX TIER 0 Tencent Hunyuan HY3 TENCENT · hy3 (open weights, Apache 2.0) via TokenHub / multiple ho… DATA GOVERNANCE UNDER REVIEW OPERATIONAL STABILITY PARTIAL ADVERSARIAL RESISTANCE UNDER REVIEW TRANSPARENCY PARTIAL COMPLIANCE POSTURE WEAK 14/19 CLAIMS QUOTE-BOUND EVIDENCE 2026-08-07
MODELRISKINDEX TIER 1 GPT-5.6 (Sol · Terra · Luna) OPENAI · gpt-5.6-sol / gpt-5.6-terra / gpt-5.6-luna via api.openai.… DATA GOVERNANCE STRONG OPERATIONAL STABILITY PARTIAL ADVERSARIAL RESISTANCE PARTIAL TRANSPARENCY STRONG COMPLIANCE POSTURE STRONG 20/21 CLAIMS QUOTE-BOUND EVIDENCE 2026-08-05
MODELRISKINDEX TIER 0 DeepSeek V4 Pro DEEPSEEK · deepseek-v4-pro (Preview) via api.deepseek.com DATA GOVERNANCE WEAK OPERATIONAL STABILITY WEAK ADVERSARIAL RESISTANCE WEAK TRANSPARENCY PARTIAL COMPLIANCE POSTURE WEAK 14/19 CLAIMS QUOTE-BOUND EVIDENCE 2026-08-05
MODELRISKINDEX TIER 0 GLM-5.2 Z.AI (ZHIPU) · glm-5.2 (open weights, MIT) via api.z.ai / self-host… DATA GOVERNANCE PARTIAL OPERATIONAL STABILITY PARTIAL ADVERSARIAL RESISTANCE WEAK TRANSPARENCY PARTIAL COMPLIANCE POSTURE WEAK 12/19 CLAIMS QUOTE-BOUND EVIDENCE 2026-08-05
MODELRISKINDEX TIER 1 NVIDIA Nemotron 3 Ultra NVIDIA · NVIDIA-Nemotron-3-Ultra-550B-A55B (open weights, OpenMDW-1… DATA GOVERNANCE NOT APPLICABLE OPERATIONAL STABILITY PARTIAL ADVERSARIAL RESISTANCE WEAK TRANSPARENCY PARTIAL COMPLIANCE POSTURE NOT APPLICABLE 12/19 CLAIMS QUOTE-BOUND EVIDENCE 2026-08-05
MODELRISKINDEX TIER 0 MiniMax M3 MINIMAX · minimax-m3 (open weights, community license) via api.mini… DATA GOVERNANCE WEAK OPERATIONAL STABILITY PARTIAL ADVERSARIAL RESISTANCE WEAK TRANSPARENCY PARTIAL COMPLIANCE POSTURE WEAK 6/17 CLAIMS QUOTE-BOUND EVIDENCE 2026-08-05
MODELRISKINDEX TIER 0 StepFun Step 3.7 Flash STEPFUN · step-3.7-flash (open weights, Apache 2.0) via platform.st… DATA GOVERNANCE WEAK OPERATIONAL STABILITY WEAK ADVERSARIAL RESISTANCE UNDER REVIEW TRANSPARENCY WEAK COMPLIANCE POSTURE WEAK 6/15 CLAIMS QUOTE-BOUND EVIDENCE 2026-08-05
MODELRISKINDEX TIER 0 Kimi K3 MOONSHOT AI · kimi-k3 via platform.kimi.ai (weights published 2026-… DATA GOVERNANCE WEAK OPERATIONAL STABILITY WEAK ADVERSARIAL RESISTANCE WEAK TRANSPARENCY PARTIAL COMPLIANCE POSTURE WEAK 8/17 CLAIMS QUOTE-BOUND EVIDENCE 2026-08-05
MODELRISKINDEX TIER 2 Claude Opus 5 ANTHROPIC · claude-opus-5 via api.anthropic.com (dateless pinned sn… DATA GOVERNANCE STRONG OPERATIONAL STABILITY PARTIAL ADVERSARIAL RESISTANCE STRONG TRANSPARENCY STRONG COMPLIANCE POSTURE STRONG 14/15 CLAIMS QUOTE-BOUND EVIDENCE 2026-08-05
MODELRISKINDEX TIER 2 Claude Sonnet 5 ANTHROPIC · claude-sonnet-5 via api.anthropic.com (dateless pinned … DATA GOVERNANCE STRONG OPERATIONAL STABILITY PARTIAL ADVERSARIAL RESISTANCE PARTIAL TRANSPARENCY STRONG COMPLIANCE POSTURE STRONG 16/17 CLAIMS QUOTE-BOUND EVIDENCE 2026-08-05
MODELRISKINDEX TIER 2 Gemini 2.5 Flash GOOGLE · gemini-2.5-flash via generativelanguage.googleapis.com DATA GOVERNANCE PARTIAL OPERATIONAL STABILITY PARTIAL ADVERSARIAL RESISTANCE WEAK TRANSPARENCY STRONG COMPLIANCE POSTURE STRONG 6/13 CLAIMS QUOTE-BOUND EVIDENCE 2026-08-05
MODELRISKINDEX TIER 0 DeepSeek V3.2 DEEPSEEK · deepseek-chat (V3.2) via api.deepseek.com DATA GOVERNANCE WEAK OPERATIONAL STABILITY PARTIAL ADVERSARIAL RESISTANCE WEAK TRANSPARENCY PARTIAL COMPLIANCE POSTURE WEAK 6/13 CLAIMS QUOTE-BOUND EVIDENCE 2026-08-05
MODELRISKINDEX TIER 2 Gemini 3.1 Pro GOOGLE · gemini-3.1-pro-preview via generativelanguage.googleapis.c… DATA GOVERNANCE PARTIAL OPERATIONAL STABILITY PARTIAL ADVERSARIAL RESISTANCE PARTIAL TRANSPARENCY STRONG COMPLIANCE POSTURE STRONG 20/23 CLAIMS QUOTE-BOUND EVIDENCE 2026-08-05
MODELRISKINDEX TIER 2 Claude Sonnet 4.5 ANTHROPIC · claude-sonnet-4-5-20250929 via api.anthropic.com DATA GOVERNANCE STRONG OPERATIONAL STABILITY PARTIAL ADVERSARIAL RESISTANCE PARTIAL TRANSPARENCY STRONG COMPLIANCE POSTURE STRONG 20/21 CLAIMS QUOTE-BOUND EVIDENCE 2026-08-05
MODELRISKINDEX TIER 1 Qwen3 235B-A22B ALIBABA (QWEN) · Qwen3-235B-A22B-Instruct (open weights, self-hoste… DATA GOVERNANCE NOT APPLICABLE OPERATIONAL STABILITY STRONG ADVERSARIAL RESISTANCE WEAK TRANSPARENCY PARTIAL COMPLIANCE POSTURE NOT APPLICABLE 2/13 CLAIMS QUOTE-BOUND EVIDENCE 2026-08-05
MODELRISKINDEX TIER 1 Llama 4 Maverick META · Llama-4-Maverick-17B-128E-Instruct (open weights, self-hoste… DATA GOVERNANCE NOT APPLICABLE OPERATIONAL STABILITY STRONG ADVERSARIAL RESISTANCE WEAK TRANSPARENCY PARTIAL COMPLIANCE POSTURE NOT APPLICABLE 8/15 CLAIMS QUOTE-BOUND EVIDENCE 2026-08-05
MODELRISKINDEX TIER 1 Kimi K2 MOONSHOT AI · Kimi-K2-Instruct (open weights, self-hosted reference) DATA GOVERNANCE NOT APPLICABLE OPERATIONAL STABILITY STRONG ADVERSARIAL RESISTANCE WEAK TRANSPARENCY PARTIAL COMPLIANCE POSTURE NOT APPLICABLE 2/13 CLAIMS QUOTE-BOUND EVIDENCE 2026-08-05
MODELRISKINDEX TIER 2 GPT-5.1 OPENAI · gpt-5.1-2025-11-13 via api.openai.com DATA GOVERNANCE STRONG OPERATIONAL STABILITY PARTIAL ADVERSARIAL RESISTANCE PARTIAL TRANSPARENCY STRONG COMPLIANCE POSTURE STRONG 22/23 CLAIMS QUOTE-BOUND EVIDENCE 2026-08-05
MODELRISKINDEX TIER 2 Claude Opus 4.5 ANTHROPIC · claude-opus-4-5-20251101 via api.anthropic.com DATA GOVERNANCE STRONG OPERATIONAL STABILITY PARTIAL ADVERSARIAL RESISTANCE PARTIAL TRANSPARENCY STRONG COMPLIANCE POSTURE STRONG 4/17 CLAIMS QUOTE-BOUND EVIDENCE 2026-08-05
MODELRISKINDEX TIER 1 Mistral Medium 3.1 MISTRAL AI · mistral-medium-2508 via api.mistral.ai DATA GOVERNANCE PARTIAL OPERATIONAL STABILITY PARTIAL ADVERSARIAL RESISTANCE WEAK TRANSPARENCY PARTIAL COMPLIANCE POSTURE PARTIAL 2/15 CLAIMS QUOTE-BOUND EVIDENCE 2026-08-05
MODELRISKINDEX TIER 1 Amazon Nova Premier AMAZON · amazon.nova-premier-v1 via bedrock-runtime.*.amazonaws.com DATA GOVERNANCE STRONG OPERATIONAL STABILITY PARTIAL ADVERSARIAL RESISTANCE PARTIAL TRANSPARENCY PARTIAL COMPLIANCE POSTURE STRONG 8/13 CLAIMS QUOTE-BOUND EVIDENCE 2026-08-05
MODELRISKINDEX TIER 2 Gemini 3 Pro GOOGLE · gemini-3-pro-preview via generativelanguage.googleapis.com… DATA GOVERNANCE PARTIAL OPERATIONAL STABILITY PARTIAL ADVERSARIAL RESISTANCE PARTIAL TRANSPARENCY STRONG COMPLIANCE POSTURE STRONG 24/30 CLAIMS QUOTE-BOUND EVIDENCE 2026-08-05
MODELRISKINDEX TIER 1 Grok 4.1 XAI · grok-4-1 via api.x.ai DATA GOVERNANCE PARTIAL OPERATIONAL STABILITY WEAK ADVERSARIAL RESISTANCE WEAK TRANSPARENCY PARTIAL COMPLIANCE POSTURE WEAK 4/12 CLAIMS QUOTE-BOUND EVIDENCE 2026-08-03