Open Ko-LLM Leaderboard

Indexed results from Open Ko-LLM Leaderboard. Each ranking retains its own metric and evaluation conditions.

ko_ifeval · prompt_level_loose_acc,none

openKoLLM · Open Ko-LLM Leaderboard · ko_ifeval · prompt_level_loose_acc,none · percent · Higher is better

  1. EleutherAI/pythia-160meleutherai15.789%

kornat_common · acc_norm,none

openKoLLM · Open Ko-LLM Leaderboard · kornat_common · acc_norm,none · percent · Higher is better

  1. EleutherAI/pythia-160meleutherai20.722%

kornat_social · A-SVA,none

openKoLLM · Open Ko-LLM Leaderboard · kornat_social · A-SVA,none · percent · Higher is better

  1. VIRNECT/llama-3-Korean-8B-r-v-0.1virnect52.515%
  2. lightblue/suzume-llama-3-8B-japaneselightblue52.349%
  3. mlabonne/Daredevil-8B-abliteratedmlabonne52.247%
  4. chlee10/T3Q-Llama3-8B-sft1.0-dpo1.0chlee1052.235%
  5. chlee10/T3Q-Llama3-8B-Inst-sft1.0chlee1052.213%
  6. v000000/L3.1-Niitorm-8B-DPO-t0.0001v00000052.153%
  7. lightblue/suzume-llama-3-8B-multilinguallightblue52.117%
  8. lightblue/suzume-llama-3-8B-multilinguallightblue52.117%
  9. Magpie-Align/Llama-3-8B-Magpie-Align-SFT-v0.3magpie-align52.06%
  10. migtissera/Llama-3-8B-Synthia-v3.5migtissera51.993%
  11. AIDX-ktds/ktdsbaseLM-v0.15-onbased-llama3.1aidx-ktds51.982%
  12. AIDX-ktds/ktdsbaseLM-v0.16-onbased-llama3.1aidx-ktds51.982%

ko_winogrande · acc,none

openKoLLM · Open Ko-LLM Leaderboard · ko_winogrande · acc,none · percent · Higher is better

  1. tanitan2/SOLAR-10.7B-Instruct-STOCK_SOLAR-tiestanitan276.953%
  2. tanitan2/SOLAR-10.7B-Instruct-T3Q-ko-solar-dpo-tiestanitan276.875%
  3. chihoonlee10/T3Q-ko-solar-dpo-v7.0chihoonlee1076.085%
  4. MoaData/Myrrh_solar_10.7b_3.0moadata76.085%
  5. chihoonlee10/T3Q-ko-solar-dpo-v3.0chihoonlee1075.138%
  6. chihoonlee10/T3Q-ko-solar-dpo-v1.0chihoonlee1074.27%
  7. hkss/hk-SOLAR-10.7B-v1.4hkss74.191%
  8. hwkwon/S-SOLAR-10.7B-v1.4hwkwon74.112%
  9. davidkim205/nox-solar-10.7b-v4davidkim20573.875%
  10. davidkim205/nox-solar-10.7b-v4davidkim20573.875%
  11. freewheelin/free-solar-evo-v0.13freewheelin73.56%
  12. gwonny/nox-solar-10.7b-v4-kolon-ITD-5-v2.0gwonny73.086%

ko_ifeval · prompt_level_strict_acc,none

openKoLLM · Open Ko-LLM Leaderboard · ko_ifeval · prompt_level_strict_acc,none · percent · Higher is better

  1. ENERGY-DRINK-LOVE/rtzr_dpo-v4-hqenergy-drink-love60.324%
  2. Qwen: Qwen2.5 7B Instructqwen60.324%
  3. kayfour/T3Q-Qwen2.5-7B-it-KOR-Safekayfour60.121%
  4. Gunulhona/Gemma-Ko-Mergegunulhona50.81%
  5. vicgalle/Configurable-Llama-3.1-8B-Instructvicgalle50.202%
  6. lightblue/suzume-llama-3-8B-multilingual-orpo-borda-top25lightblue48.583%
  7. Eurdem/Defne-llama3.1-8Beurdem48.178%
  8. v000000/L3.1-Niitorm-8B-DPO-t0.0001v00000048.178%
  9. mlabonne/Daredevil-8B-abliteratedmlabonne47.976%
  10. lightblue/suzume-llama-3-8B-multilingual-orpo-borda-top75lightblue47.166%
  11. vicgalle/Configurable-Llama-3-8B-v0.3vicgalle45.951%
  12. 4yo1/llama3-eng-ko-8b-sl54yo145.749%

ko_ifeval · inst_level_strict_acc,none

openKoLLM · Open Ko-LLM Leaderboard · ko_ifeval · inst_level_strict_acc,none · percent · Higher is better

  1. EleutherAI/pythia-160meleutherai19.288%

ko_gpqa_diamond_zeroshot · acc_norm,none

openKoLLM · Open Ko-LLM Leaderboard · ko_gpqa_diamond_zeroshot · acc_norm,none · percent · Higher is better

  1. EleutherAI/pythia-160meleutherai24.747%

ko_gsm8k · exact_match,strict-match

openKoLLM · Open Ko-LLM Leaderboard · ko_gsm8k · exact_match,strict-match · percent · Higher is better

  1. EleutherAI/pythia-160meleutherai0%

ko_eqbench · eqbench,none

openKoLLM · Open Ko-LLM Leaderboard · ko_eqbench · eqbench,none · score · Higher is better

  1. ENERGY-DRINK-LOVE/rtzr_dpo-v4-hqenergy-drink-love59.525
  2. T3Q-LLM/T3Q-LLM-sft1.0-dpo1.0t3q-llm55.307
  3. ai-human-lab/EEVE-Korean_Instruct-10.8B-expoai-human-lab53.588
  4. T3Q-LLM/T3Q-LLM1-CV-v2.0t3q-llm52.709
  5. algograp-Inc/algograpV4algograp-inc52.048
  6. T3Q-LLM/T3Q-LLM-solar10.8-sft-v1.0t3q-llm51.863
  7. T3Q-LLM/T3Q-LLM1-CV-v1.0t3q-llm51.756
  8. yanolja/EEVE-Korean-Instruct-10.8B-v1.0yanolja51.668
  9. kayfour/T3Q-gemma2-9B-it-Ko-safekayfour51.014
  10. sinjy1203/EEVE-Korean-Instruct-10.8B-v1.0-Grade-Retrievalsinjy120349.544
  11. cognitivecomputations/dolphin-2.9.3-mistral-nemo-12bcognitivecomputations48.53
  12. maywell/Mini_Synatra_SFTmaywell47.919

ko_winogrande · acc,none

openKoLLM · Open Ko-LLM Leaderboard · ko_winogrande · acc,none · percent · Higher is better

  1. EleutherAI/pythia-160meleutherai49.803%

ko_eqbench · percent_parseable,none

openKoLLM · Open Ko-LLM Leaderboard · ko_eqbench · percent_parseable,none · percent · Higher is better

  1. Phi-3-medium-128k-instructmicrosoft90.058%
  2. AALF/gemma-2-27b-it-SimPO-37K-100stepsaalf89.474%
  3. Cran-May/T.E-8.1cran-may89.474%
  4. Lambent/qwen2.5-reinstruct-alternate-lumen-14Blambent89.474%
  5. migtissera/Tess-3-Mistral-Nemo-12Bmigtissera89.474%
  6. migtissera/Tess-v2.5-Gemma-2-27B-alphamigtissera89.474%
  7. migtissera/Trinity-2-Codestral-22Bmigtissera89.474%
  8. nbeerbower/gemma2-gutenberg-27Bnbeerbower89.474%
  9. princeton-nlp/Llama-3-Instruct-8B-IPOprinceton-nlp89.474%
  10. recoilme/Gemma-2-Ataraxy-Gemmasutra-9B-slerprecoilme89.474%
  11. rombodawg/rombos_Replete-Coder-Llama3-8Brombodawg89.474%
  12. VAGOsolutions/SauerkrautLM-Phi-3-mediumvagosolutions89.474%

ko_ifeval · prompt_level_loose_acc,none

openKoLLM · Open Ko-LLM Leaderboard · ko_ifeval · prompt_level_loose_acc,none · percent · Higher is better

  1. nbeerbower/gemma2-gutenberg-27Bnbeerbower72.874%
  2. byroneverson/gemma-2-27b-it-abliteratedbyroneverson71.66%
  3. recoilme/recoilme-gemma-2-9B-v0.2recoilme69.028%
  4. arcee-ai/SuperNova-Mediusarcee-ai68.826%
  5. v000000/Qwen2.5-Lumen-14Bv00000067.611%
  6. rombodawg/Rombos-LLM-V2.6-Qwen-14brombodawg67.409%
  7. v000000/Qwen2.5-14B-Gutenberg-1e-Deltav00000067.206%
  8. v000000/Qwen2.5-14B-Gutenberg-Instruct-Slerpenov00000067.206%
  9. tanliboy/lambda-qwen2.5-14b-dpo-testtanliboy67.004%
  10. Lambent/qwen2.5-reinstruct-alternate-lumen-14Blambent66.599%
  11. Qwen/Qwen2.5-14B-Instructqwen66.397%
  12. UCLA-AGI/Gemma-2-9B-It-SPPO-Iter1ucla-agi65.992%

Models

Loading
EVIDENCE / COMPARISON

Benchmarks

Explore every benchmark

Loading

These charts contain model-linked results. The charts above also include unlinked source profiles. Compare scores within each evaluation. Model quality, usage, transparency and system performance measure different things.

Install TensorTrace