AI Model Benchmarks
Updated
Text-Based Models (LLM)
The chat models behind writing, research and everyday questions, scored on hard tests and blind head-to-head votes.
Text model leaderboard
Top 20 of 416 · ToolsGuide Score out of 1,000, where 1,000 is a flawless model
Vision Models
Models that understand images as well as text: photos, diagrams, screenshots and handwriting.
- Best overall
Highest ToolsGuide Score with images in the prompt
655 / 1,000
Vision model leaderboard
Top 20 of 167 · ToolsGuide Score out of 1,000, where 1,000 is a flawless model
Coding Models
Which model writes the better working software, from real engineering tests and developers' side-by-side votes.
Coding model leaderboard
Top 20 of 147 · ToolsGuide Score out of 1,000, where 1,000 is a flawless model
AI Agents
Models that take multi-step actions with tools: running commands, using apps and finishing tasks on their own.
Agent leaderboard
Top 20 of 53 · ToolsGuide Score out of 1,000, where 1,000 is a flawless model
Image Generation Models
Models that create pictures from a text prompt or edit an existing image, scored by how often voters prefer them.
- Best overall
gpt-image-2.5-sunburst OpenAIHighest Preference score for text to image
841 / 1,000 preference
- Best for editing gpt-image-2.5-flare OpenAI
Highest Preference score for image editing
746 / 1,000 preference
- Best value GPT Image 1.5 OpenAI
Lowest price per image in the top 20
$0.000 per image
Image model leaderboard
Top 20 of 93 · Preference score out of 1,000: 500 is a typical model, 1,000 one voters would always pick
Video Generation Models
Models that make video clips from a text prompt or bring a still image to life, scored by how often voters prefer them.
- Best text to video
gemini-omni-1.1-flash GoogleHighest Preference score for text to video
825 / 1,000 preference
- Best image to video
minimax-h3 MiniMaxHighest Preference score for image to video
751 / 1,000 preference
- Best value sora-2 OpenAI
Lowest price per minute in the top 20
$6.00 per minute
Video model leaderboard
Top 20 of 48 · Preference score out of 1,000: 500 is a typical model, 1,000 one voters would always pick