Benchmark Almanac

MastermindEval

Phi-4 53.7% leads since March 2025
Not in model cards

Models infer a hidden symbol sequence from previous guesses and feedback in Mastermind. The evaluation varies sequence length and the similarity of answer choices.

Made by
Jonas Golde and colleagues
Released
March 2025
Kind
Puzzles and chess
Level
School
Scored
Checked automatically
Answers
Multiple choice
Tasks
All public
Measures
Capability
Language
English
In and out
Text → Text
Interaction
Single turn
Refresh
One fixed set
Run by
Anyone
Strongest model tested
Phi-4 #160 on the capability index
Last model added
January 2025
Scores refreshed
November 20th, 2026
8
Models scored in its paper

Phi-4 leads with 53.7% among models released by January 2025

Best score per model from the benchmark's paper, read on November 20th, 2026. A paper’s table doesn’t add newer models. The headline averages multiple-choice accuracy across three code sizes and two types of incorrect answer. Where a model ran with several settings or harnesses, we keep its best.

The newest model with a score here came out in January 2025, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:

  1. 1 Phi-4 53.7%
  2. 2 Phi-3.5-mini 51.8%
  3. 3 Qwen2.5-7B 48.5%
  4. 4 DeepSeek-R1-Distill-Qwen-1.5B 46%
  5. 5 Llama 3.1-8B 44.7%
  6. 6 Qwen2.5-3B 40.5%
  7. 7 Llama 3.2 3B 40%
  8. 8 DeepSeek-R1-Distill-Qwen-7B 36.7%

Source: The MastermindEval paper

#ModelScoreSettingDateSource
1 Phi-4 53.7% Dec 2024 Official leaderboard
2 Phi-3.5-mini 51.8% Aug 2024 Official leaderboard
3 Qwen2.5-7B 48.5% Sep 2024 Official leaderboard
4 DeepSeek-R1-Distill-Qwen-1.5B 46% Jan 2025 Official leaderboard
5 Llama 3.1-8B 44.7% Jul 2024 Official leaderboard
6 Qwen2.5-3B 40.5% Sep 2024 Official leaderboard
7 Llama 3.2 3B 40% Sep 2024 Official leaderboard
8 DeepSeek-R1-Distill-Qwen-7B 36.7% Jan 2025 Official leaderboard

Rankings by part

We read 6 parts from the benchmark’s paper in 2 groups: random incorrect answers and close incorrect answers. Each ranks models on one part of the test. Phi-4 leads 3 of them, and Phi-3.5-mini leads the other 3. Arrows show how far a model moved from its overall rank among the models we track.

(c=2, n=4)

8 models we track have a score here.

  1. 1 Phi-4 44%
  2. 2 Phi-3.5-mini 38%
  3. 3 Qwen2.5-7B 31%
  4. 4 Llama 3.1-8B ↑1 30%
  5. 5 DeepSeek-R1-Distill-Qwen-1.5B ↓1 30%
  6. 6 DeepSeek-R1-Distill-Qwen-7B ↑2 26%
  7. 7 Llama 3.2 3B 23%
  8. 8 Qwen2.5-3B ↓2 22%

Number one over time

Lab claims count from the day the card came out, independent runs from the model’s release.

Microsoft
Benchmark paper Not tested since Mar 2025
Phi-4 53.7%

Source: The MastermindEval paper

Scores over time

All the scores we have, one dot each, so a model tested at several settings appears more than once.

Plot with other benchmarks
In the benchmark’s paper Best so far
0% 20% 40% 60% 80% 100% 2025 Near the ceiling (85%) Phi-4 Phi-3.5-mini Llama 3.1-8B 0% 20% 40% 60% 80% 100% 2025 Near the ceiling (85%) Phi-4 Llama 3.1-8B

Source: Official MastermindEval leaderboard

No lab model card reports it yet

We read the headline results of 540 launch posts from 57 labs, and none of them list this benchmark.

More in Games and long tasks

All 261 in Games and long tasks
  • ARC-AGI-3 In use Interactive game environments where agents must explore, infer goals and rules, and act efficiently without instructions. GPT-6.2 Astra 100%
  • Orak Fading LLM agents play 12 popular video games across most genres through an MCP interface. Each game scores in-game progress or outcomes, and the board ranks models by their average rank across the games. Gemini 2.5 Pro 3.90
  • MineDojo Fading Agents complete open-ended tasks in Minecraft, such as mining blocks and fighting mobs, from a large simulation suite. Programmatic checks score many tasks. Seed2.0 Pro 49%
  • TextQuests Only in system cards Agents play 25 classic interactive fiction games. The score is harmful actions per 100 steps, where lower is better. The board also reports game progress. Llama 4 Scout 7.00
  • GameWorld Only in system cards Multimodal game agents play 34 browser games across 170 tasks. Outcomes are checked from the game state. GPT-6.1 Terra 76.8%
  • Swarm Bench Only in system cards Models orchestrate agent swarms on complex tasks that benefit from coordinated decomposition and parallel execution. Kimi K3 76.3%

Data audit

9 of 10 checks pass, 1 warning. Look first at the Newest score check. The newest came 22 months ago, in January 2025, for DeepSeek-R1-Distill-Qwen-1.5B. Every benchmark

Catalog record 2 pass
  • Pass Task access All tasks are public.
  • Pass Confidence Checked against a primary source.
Scores 1 warning 3 pass 1 note
  • Pass Independent runs Scores from the benchmark’s paper (8 entries).
  • Pass Models scored 8 models we track.
  • Warning Newest score The newest came 22 months ago, in January 2025, for DeepSeek-R1-Distill-Qwen-1.5B.
  • Note Frontier models None of the top 10 has a score. The strongest tested is Phi-4 (#160).
  • Pass Headroom The best is 53.7%, under 85%, so it still separates models.
Leaderboard import 3 pass
  • Pass Importer We read the board on arxiv.org: 8 rows for models we track at the last read.
  • Pass Last read November 20th, 2026.
  • Pass Read worked The last read worked.
Consistency 1 pass 1 note
  • Note Lab claims No model has both a lab number and an independent run.
  • Pass Known issues No published audit found problems.

How we check