LiveBench
Indexed results from LiveBench. Each ranking retains its own metric and evaluation conditions.
LiveBench Coding
liveBench · LiveBench · Average score · percent
- OpenAI: o3 Mini (batch)openai82.813%
- QwQ-32Bqwen75.781%
- DeepSeek: DeepSeek V3 0324deepseek71.094%
- OpenAI: o3 Mini (batch)openai69.531%
- OpenAI: o1openai68.75%
- OpenAI: o3 Mini (batch)openai64.844%
- DeepSeek: DeepSeek V3deepseek61.719%
- Perplexity: Sonar Properplexity60.156%
- abacusai/Dracarys2-72B-Instructabacusai57.813%
- DeepSeek: DeepSeek-R1-Distill-Llama-70Bdeepseek55.469%
- OpenAI: o1openai55.469%
- NovaSky-AI/Sky-T1-32B-Previewnovasky-ai54.688%
LiveBench Instruction Following
liveBench · LiveBench · Average score · percent
- DeepSeek: DeepSeek V3 0324deepseek86.783%
- Qwen/Qwen2-72B-Instructqwen85.417%
- QwQ-32Bqwen85.3%
- OpenAI: o1openai83.15%
- OpenAI: o1openai82.9%
- OpenAI: o3 Mini (batch)openai82.5%
- OpenAI: o1openai80.867%
- OpenAI: o3 Mini (batch)openai79.85%
- Meta: Llama 4 Maverickmeta-llama79.583%
- Perplexity: Sonarperplexity79.311%
- OpenAI: o3 Mini (batch)openai78.583%
- gpt-4 · azure-openai v1106-Previewopenai77.083%
LiveBench Language
liveBench · LiveBench · Average score · percent
- Perplexity: Sonar Properplexity68.73%
- OpenAI: o1openai64.755%
- OpenAI: o1openai63.539%
- OpenAI: o1openai61.134%
- OpenAI: GPT-4o (2024-05-13)openai51.205%
- gpt-4 · azure-openai v0613openai49.856%
- OpenAI: o3 Mini (batch)openai49.459%
- OpenAI: GPT-4o (2024-11-20)openai49.019%
- DeepSeek: DeepSeek V3 0324deepseek48.688%
- gpt-4 · azure-openai v1106-Previewopenai48.576%
- QwQ-32Bqwen47.686%
- Meta: Llama 4 Maverickmeta-llama47.216%
LiveBench Overall
liveBench · LiveBench · Average score · percent
- OpenAI: o1openai67.451%
- Perplexity: Sonar Properplexity64.066%
- OpenAI: o3 Mini (batch)openai63.625%
- OpenAI: o1openai63.55%
- OpenAI: o1openai61.372%
- QwQ-32Bqwen60.788%
- DeepSeek: DeepSeek V3 0324deepseek60.106%
- OpenAI: o3 Mini (batch)openai56.669%
- DeepSeek: DeepSeek V3deepseek54.668%
- OpenAI: GPT-4o (2024-05-13)openai54.359%
- Meta: Llama 4 Maverickmeta-llama52.398%
- OpenAI: o3 Mini (batch)openai51.906%
Models
LoadingBenchmarks
Explore every benchmark
LoadingThese 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.