LLM Reference
activeGeneral

MMLU PRO

Metric: Accuracy (higher is better)Introduced: 2024

Enhanced MMLU with 10-option questions (vs 4 in original), harder problems filtered through expert review, and reduced shortcuts. Maintains coverage of 57 academic subjects.

Models ranked

105

tracked on this benchmark

Score band

91.8 – 20.0

best → lowest tracked

Snapshot trend

+7.43

Jun 7 → Jul 3 · 1 models

Leaderboard

Tracked models ranked by Accuracy (higher is better).

Compare candidates
#Model variant and provenanceScore
1
Gemini 3 Pro
Version: Not recordedHarness: Not recordedEvaluator: Not recordedObserved: Dec 11, 2025Confidence: Not recordedSource
91.8
2
Gemini 3.1 Pro Preview
Version: MMLU-Pro (accuracy)Harness: Not recordedEvaluator: Not recordedObserved: Jun 7, 2026Confidence: Not recordedSource
91.0
3
Qwen3.7-Max
Version: MMLU-Pro (accuracy)Harness: Not recordedEvaluator: Not recordedObserved: Jun 7, 2026Confidence: Not recordedSource
89.6
4
Claude Opus 4.6
Version: Not recordedHarness: Not recordedEvaluator: Not recordedObserved: Apr 19, 2026Confidence: Not recordedSource
89.1
5
Claude Opus 4.5
Version: Not recordedHarness: Not recordedEvaluator: Not recordedObserved: Apr 14, 2026Confidence: Not recordedSource
88.9
6
Gemini 3 Flash
Version: Not recordedHarness: Not recordedEvaluator: Not recordedObserved: Apr 19, 2026Confidence: Not recordedSource
88.6
7
Qwen3.6 Max Preview
Version: MMLU-Pro (accuracy)Harness: Not recordedEvaluator: Not recordedObserved: Jun 7, 2026Confidence: Not recordedSource
88.5
8
Qwen3.6-Plus
Version: MMLU-Pro (accuracy)Harness: Not recordedEvaluator: Not recordedObserved: Jun 7, 2026Confidence: Not recordedSource
88.5
9
Gemini 2.5 Flash
Version: Not recordedHarness: Not recordedEvaluator: Not recordedObserved: Jun 17, 2025Confidence: Not recordedSource
88.4
10
GPT-5.5
Version: Not recordedHarness: Not recordedEvaluator: Not recordedObserved: Apr 29, 2026Confidence: Not recordedSource
88.1
11
Qwen3.5-397B-A17B
Version: From official HuggingFace model card (accuracy)Harness: Not recordedEvaluator: Not recordedObserved: Jun 7, 2026Confidence: Not recordedSource
87.8
12
ByteDance Doubao Seed 2.0 Lite
Version: MMLU-Pro (accuracy)Harness: Not recordedEvaluator: Not recordedObserved: Jun 7, 2026Confidence: Not recordedSource
87.7
13
Seed 2.0 Lite
Version: From third-party comparison citing ByteDance official figures (accuracy%)Harness: Not recordedEvaluator: Not recordedObserved: Jun 7, 2026Confidence: Not recordedSource
87.7
14
GPT-5.4
Version: Not recordedHarness: Not recordedEvaluator: Not recordedObserved: Apr 19, 2026Confidence: Not recordedSource
87.5
15
DeepSeek V4 Pro
Version: Think Max mode (accuracy)Harness: Not recordedEvaluator: Not recordedObserved: Jun 7, 2026Confidence: Not recordedSource
87.5
16
Claude Sonnet 4.6
Version: Not recordedHarness: Not recordedEvaluator: Not recordedObserved: Apr 19, 2026Confidence: Not recordedSource
87.3
17
Kimi K2.5
Version: Thinking mode (accuracy)Harness: Not recordedEvaluator: Not recordedObserved: Jun 7, 2026Confidence: Not recordedSource
87.1
18
Grok 4
Version: Not recordedHarness: Not recordedEvaluator: Not recordedObserved: Apr 14, 2026Confidence: Not recordedSource
87.0
19
ByteDance Doubao Seed 2.0 Pro
Version: From ByteDance official launch materials (accuracy)Harness: Not recordedEvaluator: Not recordedObserved: Jun 7, 2026Confidence: Not recordedSource
87.0
20
ERNIE 5.0
Version: From official ERNIE 5 (accuracy%)Harness: Not recordedEvaluator: Not recordedObserved: Jun 7, 2026Confidence: Not recordedSource
87.0
21
Qwen3.5-122B-A10B
Version: From official HuggingFace model card (accuracy)Harness: Not recordedEvaluator: Not recordedObserved: Jun 7, 2026Confidence: Not recordedSource
86.7
22
DeepSeek V4 Flash
Version: MMLU-Pro EM, V4-Flash Think HighHarness: Not recordedEvaluator: Not recordedObserved: Jul 3, 2026Confidence: Not recordedSource
86.4
23
Gemini 2.5 Pro
Version: Not recordedHarness: Not recordedEvaluator: Not recordedObserved: Apr 14, 2026Confidence: Not recordedSource
86.2
24
Qwen3.6-27B
Version: From official HuggingFace model card (accuracy)Harness: Not recordedEvaluator: Not recordedObserved: Jun 7, 2026Confidence: Not recordedSource
86.2
25
GPT-4.5
Version: Not recordedHarness: Not recordedEvaluator: Not recordedObserved: Apr 14, 2026Confidence: Not recordedSource
86.1
Showing the top 25 of 105 tracked models. Browse all models.

How to read this benchmark

This benchmark scores models where higher is better. Use scores for directional filtering and shortlisting, not universal quality ranking; then validate pricing, context window, provider availability, and fit for your workload.

Trust this score when

  • There is a fresh timestamped snapshot (or multiple snapshots) for this benchmark.
  • The model list covers the same version family you can actually deploy today.
  • Top candidates overlap with your required routing and feature requirements.

Be cautious when

  • There is only one benchmark snapshot or the dataset appears stale.
  • The benchmark metric direction is opposite of your decision objective.
  • The score difference between options is narrow and likely within implementation variance.

Related benchmarks

MMLU
Massive Multitask Language Understanding
IFEval
Instruction-Following Evaluation
Artificial Analysis Intelligence Index

Last reviewed: Apr 15, 2026

Resources

Website