Skip to content

Benchmark

IdrockBench Uzbek IFEval (prompt-level strict, task v2.0)

Instruction following benchmarks measure how closely models follow instructions. Scores are percentages. Higher is better.

Models measured
12
Scores recorded
12
Sources
1
IQ weight
1.05

Score distribution

56.6%
Mean
64.5%
Median
17.3%
Std dev
29.7%
10th percentile
75.5%
90th percentile
27.9%
Min
27.9%76%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.05 for chat models, measured over 12 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard

#ModelScore
1Qwen3.5 27BQ4_K_M, reasoning: off, organiser-run· IdrockBench Uzbek leaderboard data docs76%
2Gemma 4 31B ITQ4_K_M, reasoning: off, organiser-run· IdrockBench Uzbek leaderboard data docs75.5%
3Qwen3.6 27BQ4_K_M, reasoning: off, organiser-run· IdrockBench Uzbek leaderboard data docs73.5%
4Gemma 4 26BQ4_K_M, reasoning: off, organiser-run· IdrockBench Uzbek leaderboard data docs70.6%
5Qwen 3.8 27BQ4_K_M, reasoning: off, organiser-run· IdrockBench Uzbek leaderboard data docs70.3%
6gemma 4 E4BQ4_K_M, reasoning: off, organiser-run· IdrockBench Uzbek leaderboard data docs64.5%
7Qwen3.5 9BQ4_K_M, reasoning: off, organiser-run· IdrockBench Uzbek leaderboard data docs56.2%
8Gemma 4 E2B ITQ4_K_M, reasoning: off, organiser-run· IdrockBench Uzbek leaderboard data docs52.8%
9Qwen3.5 4BQ4_K_M, reasoning: off, organiser-run· IdrockBench Uzbek leaderboard data docs47.6%
10Qwen3.5 2BQ8_0, reasoning: off, organiser-run· IdrockBench Uzbek leaderboard data docs34.8%
11Qwen3.5 0.8BQ8_0, reasoning: off, organiser-run· IdrockBench Uzbek leaderboard data docs29.7%
12Nemotron 3.5 Lightning 30bQ4_K_M, reasoning: off, organiser-run· IdrockBench Uzbek leaderboard data docs27.9%

Data credits: IdrockBench, IDROCK AI Excellence Center, New Uzbekistan University. Public numeric results; no results licence stated (code MIT and dataset licences not borrowed).