Benchmark Almanac

MiJaBench

Llama 3.3 70B Instruct 58.2% leads since January 2026
Not in model cards

Models respond to 43,961 English and Portuguese prompts across 16 minority groups. Judges measure how defense rates change with the target group and language.

The 2026 paper compares 14 models and identifies Gemini 2.0 Flash as its most robust baseline. Its English defense rate was 58%.

Released
January 2026
Kind
Bias, fairness and politics
Level
University
Scored
Graded by a model
Answers
Open-ended
Tasks
All public
Measures
Propensity
Risk
Hate
Behavior
Bias
Language
English, Portuguese
In and out
Text → Text
Interaction
Single turn
Refresh
One fixed set
Run by
Anyone
In views
Content safety
Strongest model tested
Qwen3-32B #135 on the capability index
Last model added
April 2025
Scores refreshed
November 20th, 2026
14
Entries on its own leaderboard, 11 models we track

Llama 3.3 70B Instruct leads with 58.2% among models released by April 2025 Llama 3.3 70B Instruct leads with 58.2% among models released by April 2025

Best score per entry from the benchmark's own leaderboard, read on November 20th, 2026. Where a model ran with several settings or harnesses, we keep its best. Names in gray aren't in our model list: fine-tunes, agent systems, products and older models. They don't count toward leaders. This view shows only models we track, each at its best.

The newest model with a score here came out in April 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 Llama 3.3 70B Instruct 58.2%
  2. 2 Gemini 2.0 Flash Feb 2025 48%
  3. 3 Llama-3.2-1B-Instruct 47.4%
  4. 4 Llama-3.1-8B-Instruct 39.1%
  5. 5 Llama-3.2-3B-Instruct 36.6%
  6. 6 GPT-4o mini 24.8%
  7. 7 Gemma 3 12B 23.2%
  8. 8 Gemma 3 27B 19.5%
  9. 9 Qwen3-32B FP8 12.6%
  10. 10 Gemma 3 4B 12%
  11. 11 Gemma 3 1B 7.18%
  12. 12 Qwen3-4B FP8 4.92%
  13. 13 Qwen3-8B FP8 4.53%
  14. 14 Qwen3-1.7B FP8 2.51%

Source: Official MiJaBench leaderboard

#EntryScoreSettingDateSource
1 Llama 3.3 70B Instruct 58.2% Dec 2024 Official leaderboard
2 Gemini 2.0 Flash Feb 2025 48% Feb 2025 Official leaderboard
3 Llama-3.2-1B-Instruct 47.4% Official leaderboard
4 Llama-3.1-8B-Instruct 39.1% Official leaderboard
5 Llama-3.2-3B-Instruct 36.6% Official leaderboard
6 GPT-4o mini 24.8% Jul 2024 Official leaderboard
7 Gemma 3 12B 23.2% Mar 2025 Official leaderboard
8 Gemma 3 27B 19.5% Mar 2025 Official leaderboard
9 Qwen3-32B 12.6% FP8 Apr 2025 Official leaderboard
10 Gemma 3 4B 12% Mar 2025 Official leaderboard
11 Gemma 3 1B 7.18% Mar 2025 Official leaderboard
12 Qwen3-4B 4.92% FP8 Apr 2025 Official leaderboard
13 Qwen3-8B 4.53% FP8 Apr 2025 Official leaderboard
14 Qwen3-1.7B 2.51% FP8 Apr 2025 Official leaderboard
#ModelScoreSettingDateSource
1 Llama 3.3 70B Instruct 58.2% Dec 2024 Official leaderboard
2 Gemini 2.0 Flash Feb 2025 48% Feb 2025 Official leaderboard
3 GPT-4o mini 24.8% Jul 2024 Official leaderboard
4 Gemma 3 12B 23.2% Mar 2025 Official leaderboard
5 Gemma 3 27B 19.5% Mar 2025 Official leaderboard
6 Qwen3-32B 12.6% FP8 Apr 2025 Official leaderboard
7 Gemma 3 4B 12% Mar 2025 Official leaderboard
8 Gemma 3 1B 7.18% Mar 2025 Official leaderboard
9 Qwen3-4B 4.92% FP8 Apr 2025 Official leaderboard
10 Qwen3-8B 4.53% FP8 Apr 2025 Official leaderboard
11 Qwen3-1.7B 2.51% FP8 Apr 2025 Official leaderboard

Rankings by part

We read 2 parts from the benchmark’s own leaderboard: English and Portuguese. Each ranks models on one part of the test. Gemini 2.0 Flash leads English and Llama 3.3 70B Instruct leads Portuguese. Arrows show how far a model moved from its overall rank among the models we track.

English

11 models we track have a score here.

  1. 1 Gemini 2.0 Flash Feb 2025 ↑1 57.6%
  2. 2 Llama 3.3 70B Instruct ↓1 47%
  3. 3 GPT-4o mini 27.6%
  4. 4 Gemma 3 12B 25.2%
  5. 5 Gemma 3 27B 23.7%
  6. 6 Qwen3-32B 16%
  7. 7 Gemma 3 4B 13.7%
  8. 8 Gemma 3 1B 7.11%
  9. 9 Qwen3-4B 5.46%
  10. 10 Qwen3-8B 4.33%

Number one over time

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

Meta
Own leaderboard Not tested since Jan 2026

Source: Official MiJaBench leaderboard

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
On the benchmark’s own leaderboard Best so far
0% 20% 40% 60% 80% 100% 2025 Near the ceiling (85%) Llama 3.3 70B Instruct GPT-4o mini 0% 20% 40% 60% 80% 100% 2025 Near the ceiling (85%) Llama 3.3 70B Instruct GPT-4o mini

Source: Official MiJaBench 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 Safety and alignment

All 636 in Safety and alignment
  • FORTRESS In use Paired adversarial and benign national-security misuse prompts. Rubric-graded risk score on adversarial and over-refusal score on benign prompts. Claude Opus 5.5 5.78%
  • StrongREJECT In use Forbidden-prompt jailbreak benchmark with a rubric grader that scores how useful a non-refusal actually is. Inkling 98.6%
  • Gray Swan IPI Arena benchmark In use Indirect prompt-injection attacks from public red-teaming competitions replayed on tool, coding and browser agents. Attack success rate at k attempts. Gemini 4 Carbon 0%
  • Covert misbehavior (Thinking Machines) In use Agentic tasks that check whether a model misbehaves and conceals it. Scored by the rate of covert misbehavior. Inkling 2 Mini 8.30%
  • Destructive actions (Thinking Machines) In use Agentic tasks that check whether a model takes destructive actions it was not authorized to take. Scored by the rate of destructive actions. Inkling 2 Small 4.70%
  • Misbehavior (Thinking Machines) In use Agentic tasks that check whether a model fabricates results, ignores instructions or hides what it did when a task is hard. Scored by the rate of misbehavior. Inkling 2 Mini 9.80%

Data audit

9 of 10 checks pass, 1 warning. Look first at the Newest score check. The newest came 19 months ago, in April 2025, for Qwen3-1.7B. 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 official leaderboard (14 entries).
  • Pass Models scored 11 models we track, plus 3 other entries on its leaderboard.
  • Warning Newest score The newest came 19 months ago, in April 2025, for Qwen3-1.7B.
  • Note Frontier models None of the top 10 has a score. The strongest tested is Qwen3-32B (#135).
  • Pass Headroom The best is 58.2%, under 85%, so it still separates models.
Leaderboard import 3 pass
  • Pass Importer We read the board on huggingface.co: 11 rows for models we track and 3 other entries 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