OpenCompass LLM

Indexed results from OpenCompass LLM. Each ranking retains its own metric and evaluation conditions.

Average

openCompass · OpenCompass LLM · Average · percent · Higher is better

  1. Z.ai: GLM 5z-ai79%
  2. Qwen: Qwen3.5 397B A17Bqwen78.4%
  3. deepseek-ai/DeepSeek-V3.2-Specialedeepseek78.2%
  4. Z.ai: GLM 4.7z-ai77.6%
  5. StepFun: Step 3.5 Flashstepfun76.9%
  6. DeepSeek: DeepSeek V3.2deepseek75.3%
  7. MoonshotAI: Kimi K2 Thinkingmoonshotai75.3%
  8. MiniMax: MiniMax M2.5minimax73.3%
  9. Z.ai: GLM 4.6z-ai73.3%
  10. OpenAI: gpt-oss-120bopenai73.2%
  11. Qwen: Qwen3 235B A22B Thinking 2507qwen71.8%
  12. MiniMax: MiniMax M2minimax69.5%

Average

openCompass · OpenCompass LLM · Average · percent · Higher is better

  1. DeepSeek: DeepSeek V4 Pro 0423deepseek91%
  2. Tencent: Hy3 previewtencent82%
  3. DeepSeek: DeepSeek V4 Flash 0423deepseek79%
  4. MoonshotAI: Kimi K2.6moonshotai74.5%
  5. MiniMax: MiniMax M3minimax72.5%

Average

openCompass · OpenCompass LLM · Average · percent · Higher is better

  1. MoonshotAI: Kimi K2.6moonshotai62.9%
  2. DeepSeek: DeepSeek V4 Flash 0423deepseek61%
  3. DeepSeek: DeepSeek V4 Pro 0423deepseek55.5%
  4. Tencent: Hy3 previewtencent54.6%
  5. MiniMax: MiniMax M3minimax54%

IFEval

openCompass · OpenCompass LLM · IFEval · percent · Higher is better

  1. StepFun: Step 3.5 Flashstepfun93.2%
  2. Z.ai: GLM 5z-ai93.2%
  3. MoonshotAI: Kimi K2 Thinkingmoonshotai92.4%
  4. deepseek-ai/DeepSeek-V3.2-Specialedeepseek91.7%
  5. Qwen: Qwen3.5 397B A17Bqwen91.5%
  6. MiniMax: MiniMax M2.5minimax91.1%
  7. MiniMax: MiniMax M2minimax90.2%
  8. OpenAI: gpt-oss-120bopenai90.2%
  9. Z.ai: GLM 4.7z-ai90.2%
  10. DeepSeek: DeepSeek V3.2deepseek89.7%
  11. Qwen: Qwen3 30B A3B Thinking 2507qwen89.7%
  12. Qwen: Qwen3 Next 80B A3B Thinkingqwen89.5%

LiveCodeBenchV6

openCompass · OpenCompass LLM · LiveCodeBenchV6 · percent · Higher is better

  1. Z.ai: GLM 5z-ai86.2%
  2. StepFun: Step 3.5 Flashstepfun83.9%
  3. Z.ai: GLM 4.7z-ai83.8%
  4. Qwen: Qwen3.5 397B A17Bqwen83%
  5. deepseek-ai/DeepSeek-V3.2-Specialedeepseek80.9%
  6. OpenAI: gpt-oss-120bopenai78.4%
  7. Z.ai: GLM 4.6z-ai78.2%
  8. MoonshotAI: Kimi K2 Thinkingmoonshotai77.1%
  9. DeepSeek: DeepSeek V3.2deepseek75.4%
  10. MiniMax: MiniMax M2minimax74%
  11. MiniMax: MiniMax M2.5minimax73.6%
  12. Qwen: Qwen3 235B A22B Thinking 2507qwen70.6%

Average

openCompass · OpenCompass LLM · Average · percent · Higher is better

  1. MoonshotAI: Kimi K2.6moonshotai75.7%
  2. Tencent: Hy3 previewtencent72.2%
  3. DeepSeek: DeepSeek V4 Pro 0423deepseek71.2%
  4. DeepSeek: DeepSeek V4 Flash 0423deepseek70.4%
  5. MiniMax: MiniMax M3minimax63%

Academic_Reasoning

openCompass · OpenCompass LLM · Academic_Reasoning · percent · Higher is better

  1. DeepSeek: DeepSeek V4 Flash 0423deepseek49.1%
  2. MoonshotAI: Kimi K2.6moonshotai47.7%
  3. DeepSeek: DeepSeek V4 Pro 0423deepseek41.3%
  4. MiniMax: MiniMax M3minimax38.1%
  5. Tencent: Hy3 previewtencent36.9%

Math

openCompass · OpenCompass LLM · Math · percent · Higher is better

  1. MoonshotAI: Kimi K2.6moonshotai75.7%
  2. Tencent: Hy3 previewtencent72.2%
  3. DeepSeek: DeepSeek V4 Pro 0423deepseek71.2%
  4. DeepSeek: DeepSeek V4 Flash 0423deepseek70.4%
  5. MiniMax: MiniMax M3minimax63%

GPQA-Diamond

openCompass · OpenCompass LLM · GPQA-Diamond · percent · Higher is better

  1. Qwen: Qwen3.5 397B A17Bqwen88.4%
  2. Z.ai: GLM 4.7z-ai86.9%
  3. deepseek-ai/DeepSeek-V3.2-Specialedeepseek86.7%
  4. Z.ai: GLM 5z-ai85.3%
  5. DeepSeek: DeepSeek V3.2deepseek84.6%
  6. MiniMax: MiniMax M2.5minimax84.6%
  7. StepFun: Step 3.5 Flashstepfun83.7%
  8. MoonshotAI: Kimi K2 Thinkingmoonshotai82.7%
  9. DeepSeek: R1 0528deepseek80.6%
  10. Z.ai: GLM 4.6z-ai80.4%
  11. Qwen: Qwen3 235B A22B Thinking 2507qwen79.8%
  12. Z.ai: GLM 4.5z-ai79.5%

Knowledge_Science

openCompass · OpenCompass LLM · Knowledge_Science · percent · Higher is better

  1. DeepSeek: DeepSeek V4 Pro 0423deepseek97.1%
  2. MiniMax: MiniMax M3minimax90.8%
  3. MoonshotAI: Kimi K2.6moonshotai90%
  4. Tencent: Hy3 previewtencent88.8%
  5. DeepSeek: DeepSeek V4 Flash 0423deepseek84.2%

MMLU-Pro

openCompass · OpenCompass LLM · MMLU-Pro · percent · Higher is better

  1. Qwen: Qwen3.5 397B A17Bqwen87.6%
  2. DeepSeek: DeepSeek V3.2deepseek85.8%
  3. deepseek-ai/DeepSeek-V3.2-Specialedeepseek85.5%
  4. Z.ai: GLM 5z-ai85.2%
  5. MoonshotAI: Kimi K2 Thinkingmoonshotai84.3%
  6. Z.ai: GLM 4.7z-ai84%
  7. DeepSeek: R1 0528deepseek83.5%
  8. Qwen: Qwen3 235B A22B Thinking 2507qwen83.5%
  9. StepFun: Step 3.5 Flashstepfun83.5%
  10. Z.ai: GLM 4.6z-ai83%
  11. Z.ai: GLM 4.5z-ai82.7%
  12. Qwen: Qwen3 Next 80B A3B Thinkingqwen82%

HLE

openCompass · OpenCompass LLM · HLE · percent · Higher is better

  1. deepseek-ai/DeepSeek-V3.2-Specialedeepseek28.6%
  2. Z.ai: GLM 5z-ai28.1%
  3. Qwen: Qwen3.5 397B A17Bqwen27.5%
  4. Z.ai: GLM 4.7z-ai25.4%
  5. DeepSeek: DeepSeek V3.2deepseek23.2%
  6. MiniMax: MiniMax M2.5minimax22.2%
  7. StepFun: Step 3.5 Flashstepfun21.6%
  8. MoonshotAI: Kimi K2 Thinkingmoonshotai21.3%
  9. Z.ai: GLM 4.6z-ai19.3%
  10. Qwen: Qwen3 235B A22B Thinking 2507qwen18.5%
  11. OpenAI: gpt-oss-120bopenai18.3%
  12. Z.ai: GLM 4.5z-ai16.9%

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