Developer
xAI
14 indexed models; 11 currently meet the evidence threshold for the overall ranking. Together they have results from 69 evaluations.
Company governance evidence
xAI is represented at -1.23 SD relative to the matched Future of Life Institute edition. Provenance: Direct. This company-level evidence contributes 10% of overall model rank.
Models by xAI
| Model | Evals | Components | Rank | Release date |
|---|---|---|---|---|
| Grok 4.20 Multi Agent | 4 | 4/7 | 42 | 2026-03-10 |
| Grok 4.5 | 21 | 7/7 | 58 | 2026-07-08 |
| Grok 4.3 | 24 | 7/7 | 115 | 2026-05-15 |
| Grok 4 Fast | 18 | 6/7 | 134 | 2025-09-19 |
| Grok 3 Mini | 17 | 7/7 | 136 | 2025-04-03 |
| Grok 4 | 33 | 7/7 | 148 | 2025-07-09 |
| Grok 4.20 | 27 | 7/7 | 151 | 2026-03-10 |
| Grok 4.6 | 13 | 6/7 | 163 | 2026-08-12 |
| Grok 3 | 19 | 7/7 | 181 | 2025-04-03 |
| Grok 4.1 Fast | 25 | 7/7 | 214 | 2025-11-19 |
| Grok 2 | 4 | 3/7 | 328 | 2024-12-12 |
| Grok 3 Beta | 2 | 3/7 | — | 2025-02-17 |
| Grok Build 0.1 | 2 | 2/7 | — | 2026-05-20 |
| Grok Code Fast 1 | 2 | 2/7 | — | 2025-08-28 |
Evaluations covering xAI models (69)
AA-Omniscience · Adversarial Humanities Benchmark (AHB) — Table 5 · Adversarial Poetry — AILuminate Baseline and Poetry ASR · AgentDrive Safety Compliance · AIRBench 2024 Safety Scenarios · Alignment Leaderboard · ANIMA · Anthropic Agentic Misalignment — blackmail · Anthropic Agentic Misalignment — corporate espionage · Anthropic Agentic Misalignment — lethal action · Arena Factuality — Search Arena (factuality-only weighting) · Arena Factuality — Text Arena (factuality-only weighting) · AuAu Authoritarian Response Audit · BioSecBench-Refusal · BioTIER · BrokenMath · BullshitBench v2 · CAIS Risk Index · CheatBench direct cheating propensity · Cisco AI Defense Rolling Single-Turn Leaderboard · Claude Fable 5.1 card — Gray Swan indirect prompt injection k=15 · Concordia AI Risk Monitor · Confabulations · DelusionEval · DystopiaBench · Emergent Collusion · Enkrypt AI Safety Leaderboard · Every Model Cheats — Cybench Cheat Propensity · FlagEval Safety and Values · Google Gemini 3.8 launch — Gray Swan indirect prompt injection k=15 · Gray Swan indirect prompt injection (15 attempts) · HELM Safety · HUMAINE Trust, Ethics and Safety · Human Pathogen Capabilities Test (HPCT) — overall refusal · kindbench v0.1.0 psychological safety ranking · LiveSecBench · MACHIAVELLI · Manager Coercion Bench · MANTA · MASK · MORU · MT-JailBench CrescendoX · NESSiE Necessary Safety Benchmark · ODCV-Bench · Olam Social Poker — Social Lie Rate · Opposite-Narrator Sycophancy · PacifAIst · PandaBench JBB direct-request panel · Pander Score · PHARE · RealityTest — Text AI-Identity Disclosure · RefusalBench · Shell · SM-Bench · Social Welfare Function Benchmark · SOSBench · SpeciesismBench · SpeciEval · StereoTales Harmful Associations · TAC · The Dictatorship Eval · TrustLLM contemporary collapsed application · TukaBench · UAVBench safety-critical decision recognition · Vals AI Cheating Audit · Vectara HHEM Factual Consistency · VETO Misfired Alignment · Vigil Mental Health Safety · WildClawBench Safety & Alignment (OpenClaw harness)