Knowledge · Benchmark profile

MMLU-Redux

A harder refresh of MMLU intended to keep broad knowledge evaluation useful after the original benchmark became too easy for frontier models.

Reference2026Active11 models

Data verified 21 Jul 2026 · Methodology 1.6.0

Benchmark score on MMLU-Redux

Leaderboard · 11 models

Claude Opus 4.5 leads at 96.6%, followed by Qwen3.7-Max (95%) and Qwen3.5 397B (94.9%).

Display onlyKnowledge2026

#1Closed

Claude Opus 4.5

Anthropic

claude-opus-4-5

96.6%

Overall #35 · Context 1M tokens

#2Closed

Qwen3.7-Max

Alibaba Cloud

qwen3.7-max

95%

Overall #37 · Context 1M tokens

#3Open

Qwen3.5 397B

Alibaba Cloud

qwen3.5-397b-a17b

94.9%

Overall #56 · Context 262K tokens

Visual analysis

See the field.

Switch between model placement, score distribution and descriptive provider averages. Every view uses the same sourced leaderboard.

ViewModelsColours

AnthropicAlibaba CloudDeepSeekJetBrainsOpenBMB

MMLU-Redux leaderboard

11 ranked models · higher is better

View accessible chart data

Accessible data for MMLU-Redux leaderboard
Model	Rank	Provider	Score
Claude Opus 4.5	#1	Anthropic	96.6%
Qwen3.7-Max	#2	Alibaba Cloud	95%
Qwen3.5 397B	#3	Alibaba Cloud	94.9%
Qwen3.6 Plus	#4	Alibaba Cloud	94.5%
Qwen3.7-Plus	#5	Alibaba Cloud	94.5%
Qwen3.6 27B	#6	Alibaba Cloud	93.5%
DeepSeek V4 Pro Base	#7	DeepSeek	90.8%
DeepSeek V4 Flash Base	#8	DeepSeek	89.4%
Mellum2-12B-A2.5B-Thinking	#9	JetBrains	86.2%
Mellum2-12B-A2.5B-Instruct	#10	JetBrains	78.1%
MiniCPM5-1B	#11	OpenBMB	70.1%

Benchmark score table (11 models)

One best score per model · higher is better

MMLU-Redux leaderboard ranked by score
Rank	Model	Provider	License	Evidence use	Score
#1	Claude Opus 4.5 claude-opus-4-5	Anthropic	closed	Reference only	96.6%
#2	Qwen3.7-Max qwen3.7-max	Alibaba Cloud	—	Reference only	95%
#3	Qwen3.5 397B qwen3.5-397b-a17b	Alibaba Cloud	open	Estimated reference	94.9%
#4	Qwen3.6 Plus qwen3-6-plus	Alibaba Cloud	closed	Reference only	94.5%
#5	Qwen3.7-Plus qwen3.7-plus	Alibaba Cloud	—	Reference only	94.5%
#6	Qwen3.6 27B qwen3.6-27b	Alibaba Cloud	open	Estimated reference	93.5%
#7	DeepSeek V4 Pro Base deepseek-v4-pro-base	DeepSeek	open	Estimated reference	90.8%
#8	DeepSeek V4 Flash Base deepseek-v4-flash-base	DeepSeek	open	Estimated reference	89.4%
#9	Mellum2-12B-A2.5B-Thinking mellum2-12b-a2-5b-thinking	JetBrains	open	Estimated reference	86.2%
#10	Mellum2-12B-A2.5B-Instruct mellum2-12b-a2-5b-instruct	JetBrains	open	Estimated reference	78.1%
#11	MiniCPM5-1B minicpm5-1b	OpenBMB	open	Estimated reference	70.1%

The top of this snapshot is led by Claude Opus 4.5 at 96.6%; third place is 1.7 points behind. The top-10 spread is 18.5 points.

Compare top models

Claude Opus 4.5 vs Qwen3.7-Max Qwen3.7-Max vs Qwen3.5 397B Qwen3.5 397B vs Qwen3.6 Plus Qwen3.6 Plus vs Qwen3.7-Plus

About MMLU-Redux

Definition and scoring

Organisation: Qwen
Category: Knowledge
Version: 2026
Direction: higher is better
Ranking use: Reference
Contamination risk: Unknown

A harder refresh of MMLU intended to keep broad knowledge evaluation useful after the original benchmark became too easy for frontier models. Results stay tied to the exact model variant and evaluation system. Multiple systems for the same model use the best published score on this page; overall Lumina scoring uses the median of ranking-eligible rows.

Open benchmark source ↗

FAQ

Quick answers

What does MMLU-Redux measure?

A harder refresh of MMLU intended to keep broad knowledge evaluation useful after the original benchmark became too easy for frontier models.

Which model scores highest on MMLU-Redux?

Claude Opus 4.5 by Anthropic currently leads with 96.6%.

How many models are evaluated on MMLU-Redux?

11 models in the LuminaBench cohort have a qualifying score on this benchmark.

Does this affect overall Lumina rank?

No. This benchmark is display-only and does not enter the overall Lumina composite.

Artificial Analysis Openness Index

2026 · 10 results · Reference

Artificial Analysis MMLU-Pro

2026 · 4 results · Reference

Massive Multitask Language Understanding

2020 · 8 results · Reference

Graduate-Level Google-Proof Q&A

2023 · 70 results · Reference

SuperGPQA: Scaling LLM Evaluation Across 285 Graduate Disciplines

2025 · 19 results · Reference

Massive Multitask Language Understanding Professional

2024 · 42 results · Reference

Knowledge · Benchmark profile

MMLU-Redux

A harder refresh of MMLU intended to keep broad knowledge evaluation useful after the original benchmark became too easy for frontier models.

Reference2026Active11 models

Data verified 21 Jul 2026 · Methodology 1.6.0

Benchmark score on MMLU-Redux

Leaderboard · 11 models

Claude Opus 4.5 leads at 96.6%, followed by Qwen3.7-Max (95%) and Qwen3.5 397B (94.9%).

Display onlyKnowledge2026

#1Closed

Claude Opus 4.5

Anthropic

claude-opus-4-5

96.6%

Overall #35 · Context 1M tokens

#2Closed

Qwen3.7-Max

Alibaba Cloud

qwen3.7-max

95%

Overall #37 · Context 1M tokens

#3Open

Qwen3.5 397B

Alibaba Cloud

qwen3.5-397b-a17b

94.9%

Overall #56 · Context 262K tokens

Visual analysis

See the field.

Switch between model placement, score distribution and descriptive provider averages. Every view uses the same sourced leaderboard.

ViewModelsColours

AnthropicAlibaba CloudDeepSeekJetBrainsOpenBMB

MMLU-Redux leaderboard

11 ranked models · higher is better

View accessible chart data

Accessible data for MMLU-Redux leaderboard
Model	Rank	Provider	Score
Claude Opus 4.5	#1	Anthropic	96.6%
Qwen3.7-Max	#2	Alibaba Cloud	95%
Qwen3.5 397B	#3	Alibaba Cloud	94.9%
Qwen3.6 Plus	#4	Alibaba Cloud	94.5%
Qwen3.7-Plus	#5	Alibaba Cloud	94.5%
Qwen3.6 27B	#6	Alibaba Cloud	93.5%
DeepSeek V4 Pro Base	#7	DeepSeek	90.8%
DeepSeek V4 Flash Base	#8	DeepSeek	89.4%
Mellum2-12B-A2.5B-Thinking	#9	JetBrains	86.2%
Mellum2-12B-A2.5B-Instruct	#10	JetBrains	78.1%
MiniCPM5-1B	#11	OpenBMB	70.1%

Benchmark score table (11 models)

One best score per model · higher is better

MMLU-Redux leaderboard ranked by score
Rank	Model	Provider	License	Evidence use	Score
#1	Claude Opus 4.5 claude-opus-4-5	Anthropic	closed	Reference only	96.6%
#2	Qwen3.7-Max qwen3.7-max	Alibaba Cloud	—	Reference only	95%
#3	Qwen3.5 397B qwen3.5-397b-a17b	Alibaba Cloud	open	Estimated reference	94.9%
#4	Qwen3.6 Plus qwen3-6-plus	Alibaba Cloud	closed	Reference only	94.5%
#5	Qwen3.7-Plus qwen3.7-plus	Alibaba Cloud	—	Reference only	94.5%
#6	Qwen3.6 27B qwen3.6-27b	Alibaba Cloud	open	Estimated reference	93.5%
#7	DeepSeek V4 Pro Base deepseek-v4-pro-base	DeepSeek	open	Estimated reference	90.8%
#8	DeepSeek V4 Flash Base deepseek-v4-flash-base	DeepSeek	open	Estimated reference	89.4%
#9	Mellum2-12B-A2.5B-Thinking mellum2-12b-a2-5b-thinking	JetBrains	open	Estimated reference	86.2%
#10	Mellum2-12B-A2.5B-Instruct mellum2-12b-a2-5b-instruct	JetBrains	open	Estimated reference	78.1%
#11	MiniCPM5-1B minicpm5-1b	OpenBMB	open	Estimated reference	70.1%

The top of this snapshot is led by Claude Opus 4.5 at 96.6%; third place is 1.7 points behind. The top-10 spread is 18.5 points.

Compare top models

Claude Opus 4.5 vs Qwen3.7-Max Qwen3.7-Max vs Qwen3.5 397B Qwen3.5 397B vs Qwen3.6 Plus Qwen3.6 Plus vs Qwen3.7-Plus

About MMLU-Redux

Definition and scoring

Organisation: Qwen
Category: Knowledge
Version: 2026
Direction: higher is better
Ranking use: Reference
Contamination risk: Unknown

Open benchmark source ↗

FAQ

Quick answers

What does MMLU-Redux measure?

A harder refresh of MMLU intended to keep broad knowledge evaluation useful after the original benchmark became too easy for frontier models.

Which model scores highest on MMLU-Redux?

Claude Opus 4.5 by Anthropic currently leads with 96.6%.

How many models are evaluated on MMLU-Redux?

11 models in the LuminaBench cohort have a qualifying score on this benchmark.

Does this affect overall Lumina rank?

No. This benchmark is display-only and does not enter the overall Lumina composite.

Artificial Analysis Openness Index

2026 · 10 results · Reference

Artificial Analysis MMLU-Pro

2026 · 4 results · Reference

Massive Multitask Language Understanding

2020 · 8 results · Reference

Graduate-Level Google-Proof Q&A

2023 · 70 results · Reference

SuperGPQA: Scaling LLM Evaluation Across 285 Graduate Disciplines

2025 · 19 results · Reference

Massive Multitask Language Understanding Professional

2024 · 42 results · Reference

Search LUMINA BENCH

Navigation

MMLU-Redux

Leaderboard · 11 models

Claude Opus 4.5

Qwen3.7-Max

Qwen3.5 397B

See the field.

MMLU-Redux leaderboard

Crop MMLU-Redux leaderboard

Benchmark score table (11 models)

Crop MMLU-Redux benchmark table

Compare top models

Definition and scoring

Quick answers

What does MMLU-Redux measure?

Which model scores highest on MMLU-Redux?

How many models are evaluated on MMLU-Redux?

Does this affect overall Lumina rank?

MMLU-Redux

Leaderboard · 11 models

Claude Opus 4.5

Qwen3.7-Max

Qwen3.5 397B

See the field.

MMLU-Redux leaderboard

Crop MMLU-Redux leaderboard

Benchmark score table (11 models)

Crop MMLU-Redux benchmark table

Compare top models

Definition and scoring

Quick answers

What does MMLU-Redux measure?

Which model scores highest on MMLU-Redux?

How many models are evaluated on MMLU-Redux?

Does this affect overall Lumina rank?

Leaderboard · 11 models

Claude Opus 4.5

Qwen3.7-Max

Qwen3.5 397B

See the field.

Benchmark score table (11 models)

Compare top models

Definition and scoring

Quick answers

What does MMLU-Redux measure?

Which model scores highest on MMLU-Redux?

How many models are evaluated on MMLU-Redux?

Does this affect overall Lumina rank?

More Knowledge benchmarks

Leaderboard · 11 models

Claude Opus 4.5

Qwen3.7-Max

Qwen3.5 397B

See the field.

Benchmark score table (11 models)

Compare top models

Definition and scoring

Quick answers

What does MMLU-Redux measure?

Which model scores highest on MMLU-Redux?

How many models are evaluated on MMLU-Redux?

Does this affect overall Lumina rank?

More Knowledge benchmarks