Compare

AI model pricing & context

34 planning tiers across 8 providers. Catalog updated 2026-09-26. Sticker prices lie without your prompt — open the token counter to weigh a real draft.

Weigh a prompt
ModelProviderInput $/1MOutput $/1MContextTokenizer
Gemini 2.0 FlashgeminiGoogle$0.1$0.41,000,000Estimate
Mistral SmallmistralMistral$0.1$0.3128,000Estimate
DeepSeek V4 FlashdeepseekDeepSeek$0.14$0.281,000,000Estimate
GPT-4o minigptOpenAI$0.15$0.6128,000Exact (browser)
Gemini 2.5 FlashgeminiGoogle$0.15$0.61,000,000Estimate
Grok 4.1 FastgrokxAI$0.2$0.52,000,000Estimate
Qwen3 235BqwenAlibaba$0.2$0.6262,144Estimate
Gemini 3.1 Flash-LitegeminiGoogle$0.25$1.51,048,576Estimate
Llama 4 MaverickllamaMeta$0.27$0.851,000,000Estimate
Qwen 2.5 72BqwenAlibaba$0.35$0.4131,072Estimate
GPT-4.1 minigptOpenAI$0.4$1.61,047,576Exact (browser)
DeepSeek V4 ProdeepseekDeepSeek$0.435$0.871,000,000Estimate
DeepSeek R1deepseekDeepSeek$0.55$2.19128,000Estimate
Llama 3.3 70BllamaMeta$0.59$0.79128,000Estimate
GPT-5.4 minigptOpenAI$0.75$4.5400,000Exact (browser)
Claude Haiku 4.5claudeAnthropic$1$5200,000Estimate
o4-minigptOpenAI$1.1$4.4200,000Exact (browser)
GPT-5gptOpenAI$1.25$10400,000Exact (browser)
Gemini 2.5 ProgeminiGoogle$1.25$101,000,000Estimate
Grok 4.20grokxAI$1.25$2.51,000,000Estimate
Grok 4.3grokxAI$1.25$2.51,000,000Estimate
GPT-4.1gptOpenAI$2$81,047,576Exact (browser)
Claude Sonnet 5claudeAnthropic$2$101,000,000Estimate
Gemini 3.1 ProgeminiGoogle$2$121,048,576Estimate
Grok 4.6grokxAI$2$6500,000Estimate
Mistral LargemistralMistral$2$6128,000Estimate
GPT-5.4gptOpenAI$2.5$15400,000Exact (browser)
GPT-4ogptOpenAI$2.5$10128,000Exact (browser)
Claude Sonnet 4.6claudeAnthropic$3$151,000,000Estimate
Claude Sonnet 4.5claudeAnthropic$3$15200,000Estimate
Claude Opus 4.7claudeAnthropic$5$251,000,000Estimate
Claude Opus 4.6claudeAnthropic$5$251,000,000Estimate
o3gptOpenAI$10$40200,000Exact (browser)
GPT-5 ProgptOpenAI$15$120400,000Exact (browser)

How to use this table

Sort mentally by job: cheap Flash/DeepSeek tiers for high-volume classification; frontier models for hard reasoning. Then paste a production prompt into Fluxkit and read the live cost chart — tokenization differs by family, so the same English can bill differently.

Rates change. Treat these as planning defaults and confirm on vendor pricing pages before you lock a budget.