Hugging Face's logo

TeichAI
/
Qwen3-4B-Thinking-2507-Gemini-2.5-Flash-Distill-GGUF

Text Generation
Transformers
GGUF
English
qwen3
text-generation-inference
unsloth
conversational
Model card Files
xet
Community

Qwen3 4B Thinking 2507 x Gemini 2.5 Flash

This model was trained on a large Gemini 2.5 Flash dataset.

The goal of was to distill the behavior, reasoning traces, output style, and (most importantly) knowledge of Gemini-2.5 Flash.

Benchmark Results

alt="Results Bar Chart"

alt="MMLU Subject Breakdown"

Model Comparison vs Base

  • Base model: unsloth/Qwen3-4B-Thinking-2507
Compare Model Benchmark Base Score Model Score Delta Delta %
TeichAI/Qwen3-4B-Thinking-2507-Gemini-2.5-Flash-Distill arc_challenge 0.486348 0.511945 0.0255973 0.0526316
TeichAI/Qwen3-4B-Thinking-2507-Gemini-2.5-Flash-Distill gpqa_diamond_zeroshot 0.30303 0.353535 0.0505051 0.166667
TeichAI/Qwen3-4B-Thinking-2507-Gemini-2.5-Flash-Distill hellaswag 0.479785 0.504382 0.0245967 0.0512661
TeichAI/Qwen3-4B-Thinking-2507-Gemini-2.5-Flash-Distill mmlu 0.65532 0.661587 0.00626691 0.00956314
TeichAI/Qwen3-4B-Thinking-2507-Gemini-2.5-Flash-Distill truthfulqa_mc2 0.555747 0.552899 -0.00284708 -0.00512299
TeichAI/Qwen3-4B-Thinking-2507-Gemini-2.5-Flash-Distill winogrande 0.64562 0.65588 0.0102605 0.0158924

Aggregate Comparison

Compare Model Benchmarks Compared Wins vs Base Ties vs Base Losses vs Base Avg Delta
TeichAI/Qwen3-4B-Thinking-2507-Gemini-2.5-Flash-Distill 6 5 0 1 0.0190632

Detailed Results

Model Benchmark Score Total Questions Total Correct
TeichAI/Qwen3-4B-Thinking-2507-Gemini-2.5-Flash-Distill arc_challenge 0.511945 1172 600
TeichAI/Qwen3-4B-Thinking-2507-Gemini-2.5-Flash-Distill gpqa_diamond_zeroshot 0.353535 198 70
TeichAI/Qwen3-4B-Thinking-2507-Gemini-2.5-Flash-Distill hellaswag 0.504382 10042 5065
TeichAI/Qwen3-4B-Thinking-2507-Gemini-2.5-Flash-Distill mmlu 0.661587 14042 9290
TeichAI/Qwen3-4B-Thinking-2507-Gemini-2.5-Flash-Distill truthfulqa_mc2 0.552899 817 451
TeichAI/Qwen3-4B-Thinking-2507-Gemini-2.5-Flash-Distill winogrande 0.65588 1267 831
unsloth/Qwen3-4B-Thinking-2507 arc_challenge 0.486348 1172 570
unsloth/Qwen3-4B-Thinking-2507 gpqa_diamond_zeroshot 0.30303 198 60
unsloth/Qwen3-4B-Thinking-2507 hellaswag 0.479785 10042 4818
unsloth/Qwen3-4B-Thinking-2507 mmlu 0.65532 14042 9202
unsloth/Qwen3-4B-Thinking-2507 truthfulqa_mc2 0.555747 817 454
unsloth/Qwen3-4B-Thinking-2507 winogrande 0.64562 1267 818

MMLU Subject Breakdown

Model Subject Benchmark Score Total Questions Total Correct
TeichAI/Qwen3-4B-Thinking-2507-Gemini-2.5-Flash-Distill formal_logic mmlu_formal_logic 0.603175 126 76
TeichAI/Qwen3-4B-Thinking-2507-Gemini-2.5-Flash-Distill high_school_european_history mmlu_high_school_european_history 0.727273 165 120
TeichAI/Qwen3-4B-Thinking-2507-Gemini-2.5-Flash-Distill high_school_us_history mmlu_high_school_us_history 0.833333 204 170
TeichAI/Qwen3-4B-Thinking-2507-Gemini-2.5-Flash-Distill high_school_world_history mmlu_high_school_world_history 0.801688 237 190
TeichAI/Qwen3-4B-Thinking-2507-Gemini-2.5-Flash-Distill international_law mmlu_international_law 0.752066 121 91
TeichAI/Qwen3-4B-Thinking-2507-Gemini-2.5-Flash-Distill jurisprudence mmlu_jurisprudence 0.777778 108 84
TeichAI/Qwen3-4B-Thinking-2507-Gemini-2.5-Flash-Distill logical_fallacies mmlu_logical_fallacies 0.797546 163 130
TeichAI/Qwen3-4B-Thinking-2507-Gemini-2.5-Flash-Distill moral_disputes mmlu_moral_disputes 0.67341 346 233
TeichAI/Qwen3-4B-Thinking-2507-Gemini-2.5-Flash-Distill moral_scenarios mmlu_moral_scenarios 0.340782 895 305
TeichAI/Qwen3-4B-Thinking-2507-Gemini-2.5-Flash-Distill philosophy mmlu_philosophy 0.694534 311 216
TeichAI/Qwen3-4B-Thinking-2507-Gemini-2.5-Flash-Distill prehistory mmlu_prehistory 0.719136 324 233
TeichAI/Qwen3-4B-Thinking-2507-Gemini-2.5-Flash-Distill professional_law mmlu_professional_law 0.45176 1534 693
TeichAI/Qwen3-4B-Thinking-2507-Gemini-2.5-Flash-Distill world_religions mmlu_world_religions 0.754386 171 129
TeichAI/Qwen3-4B-Thinking-2507-Gemini-2.5-Flash-Distill business_ethics mmlu_business_ethics 0.71 100 71
TeichAI/Qwen3-4B-Thinking-2507-Gemini-2.5-Flash-Distill clinical_knowledge mmlu_clinical_knowledge 0.70566 265 187
TeichAI/Qwen3-4B-Thinking-2507-Gemini-2.5-Flash-Distill college_medicine mmlu_college_medicine 0.693642 173 120
TeichAI/Qwen3-4B-Thinking-2507-Gemini-2.5-Flash-Distill global_facts mmlu_global_facts 0.44 100 44
TeichAI/Qwen3-4B-Thinking-2507-Gemini-2.5-Flash-Distill human_aging mmlu_human_aging 0.695067 223 155
TeichAI/Qwen3-4B-Thinking-2507-Gemini-2.5-Flash-Distill management mmlu_management 0.854369 103 88
TeichAI/Qwen3-4B-Thinking-2507-Gemini-2.5-Flash-Distill marketing mmlu_marketing 0.846154 234 198
TeichAI/Qwen3-4B-Thinking-2507-Gemini-2.5-Flash-Distill medical_genetics mmlu_medical_genetics 0.77 100 77
TeichAI/Qwen3-4B-Thinking-2507-Gemini-2.5-Flash-Distill miscellaneous mmlu_miscellaneous 0.779055 783 610
TeichAI/Qwen3-4B-Thinking-2507-Gemini-2.5-Flash-Distill nutrition mmlu_nutrition 0.718954 306 220
TeichAI/Qwen3-4B-Thinking-2507-Gemini-2.5-Flash-Distill professional_accounting mmlu_professional_accounting 0.528369 282 149
TeichAI/Qwen3-4B-Thinking-2507-Gemini-2.5-Flash-Distill professional_medicine mmlu_professional_medicine 0.672794 272 183
TeichAI/Qwen3-4B-Thinking-2507-Gemini-2.5-Flash-Distill virology mmlu_virology 0.524096 166 87
TeichAI/Qwen3-4B-Thinking-2507-Gemini-2.5-Flash-Distill econometrics mmlu_econometrics 0.605263 114 69
TeichAI/Qwen3-4B-Thinking-2507-Gemini-2.5-Flash-Distill high_school_geography mmlu_high_school_geography 0.823232 198 163
TeichAI/Qwen3-4B-Thinking-2507-Gemini-2.5-Flash-Distill high_school_government_and_politics mmlu_high_school_government_and_politics 0.865285 193 167
TeichAI/Qwen3-4B-Thinking-2507-Gemini-2.5-Flash-Distill high_school_macroeconomics mmlu_high_school_macroeconomics 0.720513 390 281
TeichAI/Qwen3-4B-Thinking-2507-Gemini-2.5-Flash-Distill high_school_microeconomics mmlu_high_school_microeconomics 0.802521 238 191
TeichAI/Qwen3-4B-Thinking-2507-Gemini-2.5-Flash-Distill high_school_psychology mmlu_high_school_psychology 0.86789 545 473
TeichAI/Qwen3-4B-Thinking-2507-Gemini-2.5-Flash-Distill human_sexuality mmlu_human_sexuality 0.717557 131 94
TeichAI/Qwen3-4B-Thinking-2507-Gemini-2.5-Flash-Distill professional_psychology mmlu_professional_psychology 0.658497 612 403
TeichAI/Qwen3-4B-Thinking-2507-Gemini-2.5-Flash-Distill public_relations mmlu_public_relations 0.690909 110 76
TeichAI/Qwen3-4B-Thinking-2507-Gemini-2.5-Flash-Distill security_studies mmlu_security_studies 0.726531 245 178
TeichAI/Qwen3-4B-Thinking-2507-Gemini-2.5-Flash-Distill sociology mmlu_sociology 0.80597 201 162
TeichAI/Qwen3-4B-Thinking-2507-Gemini-2.5-Flash-Distill us_foreign_policy mmlu_us_foreign_policy 0.82 100 82
TeichAI/Qwen3-4B-Thinking-2507-Gemini-2.5-Flash-Distill abstract_algebra mmlu_abstract_algebra 0.57 100 56
TeichAI/Qwen3-4B-Thinking-2507-Gemini-2.5-Flash-Distill anatomy mmlu_anatomy 0.6 135 81
TeichAI/Qwen3-4B-Thinking-2507-Gemini-2.5-Flash-Distill astronomy mmlu_astronomy 0.809211 152 123
TeichAI/Qwen3-4B-Thinking-2507-Gemini-2.5-Flash-Distill college_biology mmlu_college_biology 0.8125 144 117
TeichAI/Qwen3-4B-Thinking-2507-Gemini-2.5-Flash-Distill college_chemistry mmlu_college_chemistry 0.51 100 51
TeichAI/Qwen3-4B-Thinking-2507-Gemini-2.5-Flash-Distill college_computer_science mmlu_college_computer_science 0.57 100 56
TeichAI/Qwen3-4B-Thinking-2507-Gemini-2.5-Flash-Distill college_mathematics mmlu_college_mathematics 0.53 100 53
TeichAI/Qwen3-4B-Thinking-2507-Gemini-2.5-Flash-Distill college_physics mmlu_college_physics 0.509804 102 52
TeichAI/Qwen3-4B-Thinking-2507-Gemini-2.5-Flash-Distill computer_security mmlu_computer_security 0.78 100 78
TeichAI/Qwen3-4B-Thinking-2507-Gemini-2.5-Flash-Distill conceptual_physics mmlu_conceptual_physics 0.753191 235 177
TeichAI/Qwen3-4B-Thinking-2507-Gemini-2.5-Flash-Distill electrical_engineering mmlu_electrical_engineering 0.724138 145 105
TeichAI/Qwen3-4B-Thinking-2507-Gemini-2.5-Flash-Distill elementary_mathematics mmlu_elementary_mathematics 0.642857 378 243
TeichAI/Qwen3-4B-Thinking-2507-Gemini-2.5-Flash-Distill high_school_biology mmlu_high_school_biology 0.835484 310 259
TeichAI/Qwen3-4B-Thinking-2507-Gemini-2.5-Flash-Distill high_school_chemistry mmlu_high_school_chemistry 0.669951 203 136
TeichAI/Qwen3-4B-Thinking-2507-Gemini-2.5-Flash-Distill high_school_computer_science mmlu_high_school_computer_science 0.82 100 82
TeichAI/Qwen3-4B-Thinking-2507-Gemini-2.5-Flash-Distill high_school_mathematics mmlu_high_school_mathematics 0.477778 270 129
TeichAI/Qwen3-4B-Thinking-2507-Gemini-2.5-Flash-Distill high_school_physics mmlu_high_school_physics 0.589404 151 89
TeichAI/Qwen3-4B-Thinking-2507-Gemini-2.5-Flash-Distill high_school_statistics mmlu_high_school_statistics 0.703704 216 152
TeichAI/Qwen3-4B-Thinking-2507-Gemini-2.5-Flash-Distill machine_learning mmlu_machine_learning 0.455357 112 51
unsloth/Qwen3-4B-Thinking-2507 formal_logic mmlu_formal_logic 0.595238 126 75
unsloth/Qwen3-4B-Thinking-2507 high_school_european_history mmlu_high_school_european_history 0.727273 165 120
unsloth/Qwen3-4B-Thinking-2507 high_school_us_history mmlu_high_school_us_history 0.818627 204 167
unsloth/Qwen3-4B-Thinking-2507 high_school_world_history mmlu_high_school_world_history 0.797468 237 189
unsloth/Qwen3-4B-Thinking-2507 international_law mmlu_international_law 0.727273 121 88
unsloth/Qwen3-4B-Thinking-2507 jurisprudence mmlu_jurisprudence 0.777778 108 84
unsloth/Qwen3-4B-Thinking-2507 logical_fallacies mmlu_logical_fallacies 0.754601 163 123
unsloth/Qwen3-4B-Thinking-2507 moral_disputes mmlu_moral_disputes 0.67341 346 233
unsloth/Qwen3-4B-Thinking-2507 moral_scenarios mmlu_moral_scenarios 0.372067 895 333
unsloth/Qwen3-4B-Thinking-2507 philosophy mmlu_philosophy 0.672026 311 208
unsloth/Qwen3-4B-Thinking-2507 prehistory mmlu_prehistory 0.70679 324 229
unsloth/Qwen3-4B-Thinking-2507 professional_law mmlu_professional_law 0.441982 1534 678
unsloth/Qwen3-4B-Thinking-2507 world_religions mmlu_world_religions 0.777778 171 133
unsloth/Qwen3-4B-Thinking-2507 business_ethics mmlu_business_ethics 0.63 100 63
unsloth/Qwen3-4B-Thinking-2507 clinical_knowledge mmlu_clinical_knowledge 0.716981 265 190
unsloth/Qwen3-4B-Thinking-2507 college_medicine mmlu_college_medicine 0.693642 173 120
unsloth/Qwen3-4B-Thinking-2507 global_facts mmlu_global_facts 0.36 100 36
unsloth/Qwen3-4B-Thinking-2507 human_aging mmlu_human_aging 0.713004 223 159
unsloth/Qwen3-4B-Thinking-2507 management mmlu_management 0.864078 103 89
unsloth/Qwen3-4B-Thinking-2507 marketing mmlu_marketing 0.854701 234 200
unsloth/Qwen3-4B-Thinking-2507 medical_genetics mmlu_medical_genetics 0.8 100 80
unsloth/Qwen3-4B-Thinking-2507 miscellaneous mmlu_miscellaneous 0.776501 783 608
unsloth/Qwen3-4B-Thinking-2507 nutrition mmlu_nutrition 0.712418 306 218
unsloth/Qwen3-4B-Thinking-2507 professional_accounting mmlu_professional_accounting 0.549645 282 155
unsloth/Qwen3-4B-Thinking-2507 professional_medicine mmlu_professional_medicine 0.669118 272 182
unsloth/Qwen3-4B-Thinking-2507 virology mmlu_virology 0.518072 166 86
unsloth/Qwen3-4B-Thinking-2507 econometrics mmlu_econometrics 0.614035 114 70
unsloth/Qwen3-4B-Thinking-2507 high_school_geography mmlu_high_school_geography 0.777778 198 154
unsloth/Qwen3-4B-Thinking-2507 high_school_government_and_politics mmlu_high_school_government_and_politics 0.829016 193 160
unsloth/Qwen3-4B-Thinking-2507 high_school_macroeconomics mmlu_high_school_macroeconomics 0.720513 390 281
unsloth/Qwen3-4B-Thinking-2507 high_school_microeconomics mmlu_high_school_microeconomics 0.802521 238 191
unsloth/Qwen3-4B-Thinking-2507 high_school_psychology mmlu_high_school_psychology 0.851376 545 464
unsloth/Qwen3-4B-Thinking-2507 human_sexuality mmlu_human_sexuality 0.740458 131 97
unsloth/Qwen3-4B-Thinking-2507 professional_psychology mmlu_professional_psychology 0.669935 612 409
unsloth/Qwen3-4B-Thinking-2507 public_relations mmlu_public_relations 0.672727 110 74
unsloth/Qwen3-4B-Thinking-2507 security_studies mmlu_security_studies 0.706122 245 173
unsloth/Qwen3-4B-Thinking-2507 sociology mmlu_sociology 0.800995 201 161
unsloth/Qwen3-4B-Thinking-2507 us_foreign_policy mmlu_us_foreign_policy 0.82 100 82
unsloth/Qwen3-4B-Thinking-2507 abstract_algebra mmlu_abstract_algebra 0.55 100 55
unsloth/Qwen3-4B-Thinking-2507 anatomy mmlu_anatomy 0.644444 135 87
unsloth/Qwen3-4B-Thinking-2507 astronomy mmlu_astronomy 0.782895 152 119
unsloth/Qwen3-4B-Thinking-2507 college_biology mmlu_college_biology 0.763889 144 110
unsloth/Qwen3-4B-Thinking-2507 college_chemistry mmlu_college_chemistry 0.55 100 55
unsloth/Qwen3-4B-Thinking-2507 college_computer_science mmlu_college_computer_science 0.62 100 62
unsloth/Qwen3-4B-Thinking-2507 college_mathematics mmlu_college_mathematics 0.44 100 44
unsloth/Qwen3-4B-Thinking-2507 college_physics mmlu_college_physics 0.480392 102 49
unsloth/Qwen3-4B-Thinking-2507 computer_security mmlu_computer_security 0.73 100 73
unsloth/Qwen3-4B-Thinking-2507 conceptual_physics mmlu_conceptual_physics 0.723404 235 170
unsloth/Qwen3-4B-Thinking-2507 electrical_engineering mmlu_electrical_engineering 0.751724 145 108
unsloth/Qwen3-4B-Thinking-2507 elementary_mathematics mmlu_elementary_mathematics 0.645503 378 244
unsloth/Qwen3-4B-Thinking-2507 high_school_biology mmlu_high_school_biology 0.812903 310 252
unsloth/Qwen3-4B-Thinking-2507 high_school_chemistry mmlu_high_school_chemistry 0.660099 203 134
unsloth/Qwen3-4B-Thinking-2507 high_school_computer_science mmlu_high_school_computer_science 0.81 100 81
unsloth/Qwen3-4B-Thinking-2507 high_school_mathematics mmlu_high_school_mathematics 0.440741 270 119
unsloth/Qwen3-4B-Thinking-2507 high_school_physics mmlu_high_school_physics 0.569536 151 86
unsloth/Qwen3-4B-Thinking-2507 high_school_statistics mmlu_high_school_statistics 0.652778 216 141
unsloth/Qwen3-4B-Thinking-2507 machine_learning mmlu_machine_learning 0.428571 112 48

Configuration

  • Quantization: 4bit
  • Temperature: 0.6
  • Top P: 0.95
  • Top K: 20
  • Repetition Penalty: 1.1

This model was finetuned and converted to GGUF format using Unsloth.

Example usage:

  • For text only LLMs: llama-cli --hf repo_id/model_name -p "why is the sky blue?"
  • For multimodal models: llama-mtmd-cli -m model_name.gguf --mmproj mmproj_file.gguf

Ollama

An Ollama Modelfile is included for easy deployment.

Downloads last month
2,139
GGUF
Model size
4B params
Architecture
qwen3
Hardware compatibility
Log In to view the estimation

3-bit

4-bit

8-bit

16-bit

Inference Providers NEW
Text Generation
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for TeichAI/Qwen3-4B-Thinking-2507-Gemini-2.5-Flash-Distill-GGUF

Base model

Qwen/Qwen3-4B-Thinking-2507
Finetuned
unsloth/Qwen3-4B-Thinking-2507
Finetuned
TeichAI/Qwen3-4B-Thinking-2507-Gemini-2.5-Flash-Distill
Quantized
(3)
this model

Dataset used to train TeichAI/Qwen3-4B-Thinking-2507-Gemini-2.5-Flash-Distill-GGUF