EQ-Bench 4
We show this table for reference; we do not rank on it.
A multi-turn benchmark of emotional and social intelligence using synthetic personas and pairwise LLM judging.
EQ-Bench 4 Elo on EQ-Bench 4 — 2026-10-24
We mirror the published eq-bench 4 elo view for EQ-Bench 4. Claude Opus 5 leads the public snapshot at 1364.9, followed by Claude Opus 5.5 (1339.5) and Muse Wonder (1324.9). We do not use these results to rank models overall.
Claude Opus 5
Anthropic
Claude Opus 5.5
Anthropic
Muse Wonder
Meta
32 modelsAgenticCurrentDisplay onlyUpdated 2026-10-24
EQ-Bench 4 Elo table (32 models)
| Rank | Model / configuration | Score | Parameters (B) | Open / closed |
|---|---|---|---|---|
| 1 | Claude Opus 5Anthropic | 1364.9 | Not reported | Closed |
| 2 | Claude Opus 5.5Anthropic | 1339.5 | Not reported | Closed |
| 3 | Muse WonderMeta | 1324.9 | Not reported | Closed |
| 4 | Kimi K3Moonshot AI | 1319.2 | Not reported | Open |
| 5 | Claude Fable 5Anthropic | 1312.5 | Not reported | Closed |
| 6 | GPT-5.5OpenAI | 1295.4 | Not reported | Closed |
| 7 | Claude Opus 4.7Anthropic | 1293.2 | Not reported | Closed |
| 8 | GPT-6.1 SolOpenAI | 1272.4 | Not reported | Closed |
| 9 | Claude Opus 4.8Anthropic | 1267.8 | Not reported | Closed |
| 10 | GPT-5.4OpenAI | 1257.7 | Not reported | Closed |
| 11 | Muse Spark 1.1Meta | 1244.4 | Not reported | Closed |
| 12 | Gemini 4 ArgonGoogle | 1236.3 | Not reported | Closed |
| 13 | GPT-5.6 SolOpenAI | 1236.1 | Not reported | Closed |
| 14 | Claude Sonnet 5Anthropic | 1221.2 | Not reported | Closed |
| 15 | GPT-5.6 TerraOpenAI | 1220.1 | Not reported | Closed |
| 16 | InklingThinking Machines Lab | 1212.1 | Not reported | Open |
| 17 | Claude Opus 4.6Anthropic | 1210.0 | Not reported | Closed |
| 18 | GLM-5.2Z.AI | 1208.1 | Not reported | Open |
| 19 | MiMo-V2.5-ProXiaomi | 1197.7 | Not reported | Closed |
| 20 | Claude Sonnet 4.6Anthropic | 1196.8 | Not reported | Closed |
| 21 | Kimi K2.6Moonshot AI | 1191.4 | Not reported | Open |
| 22 | DeepSeek V4 Pro 0813DeepSeek | 1156.3 | Not reported | Open |
| 23 | GPT-5.6 LunaOpenAI | 1145.9 | Not reported | Closed |
| 24 | MiniMax M3MiniMax | 1140.1 | Not reported | Open |
| 25 | Gemini 3.1 Pro PreviewGoogle | 1132.3 | Not reported | Unknown |
| 26 | Gemma 4 31B ITGoogle | 1110.0 | Not reported | Unknown |
| 27 | Qwen3.7 MaxAlibaba | 1100.0 | Not reported | Closed |
| 28 | Gemini 3.5 FlashGoogle | 1077.1 | Not reported | Closed |
| 29 | Grok 4.3xAI | 1064.9 | Not reported | Closed |
| 30 | Claude Haiku 4.5Anthropic | 1053.4 | Not reported | Closed |
| 31 | Qwen3.6-27BAlibaba | 1015.3 | Not reported | Open |
| 32 | Mistral Medium 3.5Mistral | 983.1 | Not reported | Unknown |
How EQ-Bench 4 is shown here
We mirror the official EQ-Bench 4 snapshot generated on 2026-10-24. Each model plays up to 120 personas through 16-turn conversations. Three models judge the transcripts across six emotional and social ability dimensions. The source reports 119 scenarios for three models; the other rows use 120.
EQ-Bench 4 reports pairwise Elo. The benchmark uses synthetic personas and model judges without human validation. We preserve the public scores and confidence intervals as display-only behavioral evidence.
Snapshot
The published EQ-Bench 4 snapshot places Claude Opus 5 first at 1364.9. The third row is 40.0 score units behind. The broader top-10 range is 107.2 score units, so the table still separates the published systems.
32 models have been evaluated on EQ-Bench 4. The benchmark falls in the Agentic category. EQ-Bench 4 is currently displayed for reference but excluded from the scoring formula, so it does not directly affect overall rankings.
About EQ-Bench 4
Year
2026
Tasks
Up to 120 multi-turn persona scenarios per model
Format
Pairwise Elo
Difficulty
Emotional and social intelligence
EQ-Bench 4 uses up to 120 personas for 16 turns and three judge models to compare transcripts across six ability dimensions. The July 26, 2026 source snapshot includes three models with 119 scenarios. We mirror the pairwise Elo scores and confidence intervals for reference. The personas and judgments are model-generated and have not been validated against human ratings.
Freshness and provenance
Version
EQ-Bench 4 2026
Refresh cadence
Quarterly
Staleness state
Current
Question availability
Public benchmark set
BenchLM uses freshness metadata to decide whether a benchmark should still be treated as a strong differentiator, a benchmark to watch, or a display-only reference. For the full scoring policy, see the BenchLM methodology page.
Questions
What does EQ-Bench 4 measure?
A multi-turn benchmark of emotional and social intelligence using synthetic personas and pairwise LLM judging.
Which model leads the published EQ-Bench 4 snapshot?
Claude Opus 5 currently leads the published EQ-Bench 4 snapshot with 1364.9 eq-bench 4 elo. BenchLM shows this benchmark for display only and does not use it in overall rankings.
How many models are evaluated on EQ-Bench 4?
The 2026-10-24 snapshot contains 32 AI models.
Know when it’s worth switching models
The model to choose, the cheaper alternative, and the release we would wait on.
Read a sample issueJoin 7,000+ subscribers.
One email each week. Unsubscribe anytime.