Join our newsletter to receive AI model updates

*Newsletter emailed near the beginning of every month

AI Model Comparison (October 2025)

Compare major AI models supported in Cursor, including benchmarks, pricing, and capabilities.

Model
SWE-bench
LiveCodeBench
Input
(/1M)
Cached In
(/1M)
Output
(/1M)
Context
Modalities
Latency (s)
Anthropicclaude-4.1-opus74.5%74.0% (Thinking)$15.00$1.50$75.00200kText, Image1.60s
Anthropicclaude-4.5-sonnet77.2% (82.0% with parallel compute)84.5%$3.00$0.30$15.00200kText, Image0.43s
Anthropicclaude-3.5-haiku49.2%53.2%$0.80$0.08$4.00200kText, Image0.60s
OpenAIgpt-574.9%75.3% (Thinking High)$1.25$0.125$10.00400kText, Image, Audio9.98s
OpenAIgpt-5-codex74.5%Not publicly available$1.25$0.125$10.00400kText, Image, Code9.98s
OpenAIgpt-5-mini59.8%77.4% (High)$0.25$0.025$2.00400kText, Image4.53s
OpenAIgpt-5-nano34.8%Not publicly available$0.05$0.005$0.40400kText3.13s
xAIgrok-472% (estimate; not publicly available)79.0%$3.00$0.30$15.00256kText, Image, Web1.80s
xAIgrok-4-fast-reasoning70% (estimate; not publicly available)80.0%$0.20$0.02$0.502MText, Image, Web3.64s
xAIgrok-4-fast-non-reasoning65% (estimate; not publicly available)Not publicly available$0.20$0.02$0.502MText, Image, Web3.64s
xAIgrok-code-fast-170.8%Not publicly available$0.20$0.02$1.50256kText, Code2.50s
Googlegemini-2.5-pro63.8%69.0% (UI: 1/1/2025-5/1/2025)$1.25$0.31$10.002MText, Image, Video, Audio2.00s
Googlegemini-2.5-flash60.4%55.4% (Thinking)$0.30$0.075$2.501MText, Image, Video0.80s
Deepseekdeepseek-r1Not applicable73.3% (0528 variant)$0.55$0.14$2.19128kText3.00s
Deepseekdeepseek-v366%74.8% (V3.1 Thinking)$0.27$0.07$1.10128kText1.50s
“—” means no reliable public data as of October 2025.
Tip: Use search to narrow down options. Click column headers to sort. Results are sorted by the selected column.