modelbenchmark.io

What a cache hit saves

Prompt caching prices

What a cache hit saves, and how many reuses pay for the write. Every figure is derived from the prices a host lists. Nothing here is measured. All prices.

Price

1,368 models

DeepSeek V4 Flash 0731 (Nebius)Alibaba$0.14$0.14$0.140.0%never
GLM-5.3Alibaba$1.40free$0.2681.4%0
GLM-5.2Alibaba$1.40free$0.2681.4%0
GLM-5.3-FlashZhipu$0.15free$0.0380%0
DeepSeek V4 Pro 0813 (Nebius)Alibaba$1.32free$1.320.0%never
Kimi K3Kimi$3.00$3.00$0.3090%0
DeepSeek V4.1 Flash (Nebius)Alibaba$0.30free$0.300.0%never
Kimi K2.6Kimi$0.95free$0.1683.2%0
Kimi K2.7 CodeKimi$0.95free$0.1980%0
GPT OSS 120B (IONOS)OpenAI$0.1676$0.15$0.1510.5%0
MiniMax-M3MiniMax$0.30$0.90+200%$0.0680%3
Claude Haiku 4.5 (latest)Anthropic$1.00$1.25+25%$0.1090%1
Claude Sonnet 4.6Anthropic$3.00$3.75+25%$0.3090%1
GLM-5.1Zhipu$1.40free$0.2681.4%0
GPT-5.5OpenAI$5.00$5.00$0.5090%0
Claude Opus 4.8Anthropic$5.00$6.25+25%$0.5090%1
GPT-5.6 LunaOpenAI$0.20$0.25+25.0%$0.0290%1
GPT-5.6 SolOpenAI$4.00$5.00+25%$0.4090%1
Claude Fable 5Anthropic$10.00$12.50+25%$1.0090%1
Claude Sonnet 5Anthropic$2.00$2.50+25%$0.2090%1
Claude Opus 4.7Anthropic$5.00$6.25+25%$0.5090%1
GPT-5.4OpenAI$2.50free$0.2590%0
Qwen3.8 27BAlibaba$0.50$0.625+25%$0.1080%1
Claude Opus 5Anthropic$5.00$6.25+25%$0.5090%1
GPT OSS 20B (Groq)OpenAI$0.075$0.07$0.037550%0
GPT-5.6 TerraOpenAI$2.00$2.50+25%$0.2090%1
MiniMax-M2.5MiniMax$0.30free$0.0390%0
MiniMax-M2.7MiniMax$0.30$0.375+25%$0.0680%1
Qwen3.8 Max 0902Alibaba$2.00$2.50+25%$0.2587.5%1
Claude Sonnet 4.5 (latest)Anthropic$3.00$3.75+25%$0.3090%1
Gemini 3.1 Pro Preview (Vertex AI)Google$2.00$0.375$0.2090%0
GPT-6 LunaOpenAI$0.10$0.125+25.0%$0.0190%1
Kimi K2.5 (TensorX)Kimi$0.50free$0.12575%0
Claude Opus 4.6Anthropic$5.00$6.25+25%$0.5090%1
Gemini 3.5 Flash (Vertex AI)Google$1.50$0.0833$0.1590%0
Gemma 4 31B ITGoogle$0.38—$0.0586.8%—
GLM-5Alibaba$1.00free$0.2080%0
GPT-6 AstraOpenAI$10.00$12.50+25%$1.0090%1
Claude Fable 5.1Anthropic$10.00$12.50+25%$0.2597.5%1
Claude Opus 4.5 (latest)Anthropic$5.00$6.25+25%$0.5090%1
Claude Opus 5.5Anthropic$4.00$5.00+25%$0.2095%1
GPT-5.4 miniOpenAI$0.75$0.75$0.07590%0
Grok 4.6xAI$2.00$2.00$0.5075%0
Claude Haiku 5.5Anthropic$0.10$0.125+25.0%$0.0190%1
Claude Sonnet 5.5Anthropic$2.00$2.50+25%$0.1095%1
Gemini 3.8 Flash (Vertex AI)Google$1.50$0.0833$0.1590%0
Gemini 2.5 FlashGoogle$0.30$1.00+233.3%$0.0390%3
Gemini 3.7 Flash (Vertex AI)Google$1.50$0.0833$0.1590%0
GPT-6 SolOpenAI$2.00$2.50+25%$0.2090%1
GPT-6.1 SolOpenAI$2.00$2.50+25%$0.1095%1
DeepSeek R1 0528 (TensorX)DeepSeek$0.66$0.825+25.0%$0.16575%1
Gemini 3.6 Flash (Vertex AI)Google$0.75$0.0417$0.07590%0
GPT-5OpenAI$1.25$1.25$0.12590%0
GPT-5 MiniOpenAI$0.25$0.25$0.02590%0
GPT-5.4 nanoOpenAI$0.20$0.20$0.0290%0
Gemini 3 Flash Preview (Vertex AI)Google$0.50$0.0833$0.0590%0
GLM-4.7Zhipu$0.60free$0.1181.7%0
GPT-5.2OpenAI$1.75$1.75$0.17590%0
Grok 4.7xAI$2.00$0.50$0.5075%0
Qwen3.8 FlashAlibaba$0.15$0.20+33.3%$0.01689.3%1
Gemini 2.5 ProGoogle$1.25$4.50+260%$0.12590%3
Grok 4.3xAI$1.25$1.25$0.2084%0
Qwen3.7 MaxAlibaba$2.50free$2.500.0%never
Qwen3.7 PlusAlibaba$0.40free$0.400.0%never
Qwen3.6 35B A3BAlibaba$0.07free$0.03550%0
Gemini 3.1 Flash Lite (Vertex AI)Google$0.25$0.0833$0.02590%0
Gemini 3.5 Flash Lite (Vertex AI)Google$0.30$0.0833$0.0390%0
Qwen3.5 397B-A17BAlibaba$0.172—$0.1720.0%—
GLM-4.6Zhipu$0.60free$0.1181.7%0
GPT-5.3 CodexOpenAI$1.75$1.75$0.17590%0
Gemini 4 Argon (Vertex AI)Google$4.00$0.375$0.2095%0
GPT-5 NanoOpenAI$0.05$0.05$0.00590%0
GPT-5.1OpenAI$1.25$1.25$0.12590%0
Llama-3.3-70B-Instruct (IONOS)Meta$0.7262—$0.2072.5%—
GPT-4.1OpenAI$2.00free$0.5075%0
Grok 4.5xAI$2.00$2.00$0.3085%0
MiniMax-M3.1-FlashMiniMax$0.075free$0.007590%0
GPT-4.1 miniOpenAI$0.40free$0.1075%0
Qwen3.6 27BAlibaba$0.4126free$0.41260.0%never
MiMo V2.6 Flash—$0.14free$0.140.0%never
MiMo V2.6 Pro—$0.435free$0.4350.0%never
GPT-4o (2024-08-06)OpenAI$2.50$1.38$1.2550%0
DeepSeek-V3 (Deep Infra)DeepSeek$0.32$0.375+17.2%$0.13557.8%1
Gemini 2.5 Flash-LiteGoogle$0.10$1.00+900%$0.0190%10
Gemma 4 26B A4B ITGoogle$0.144—$0.0565.3%—
GPT-4o miniOpenAI$0.15$0.15$0.07550%0
GPT-6.1 TerraOpenAI$1.00$1.25+25%$0.0595%1
MiniMax-M2.1MiniMax$0.30$0.375+25%$0.0390%1
Mistral Large 3Mistral$0.50—$0.0590%—
Qwen3-Next 80B-A3B Instruct (Deep Infra)Alibaba$0.09$0.15+66.7%$0.01583.3%1
Qwen3.6 PlusAlibaba$0.50free$0.500.0%never
GLM-4.5-AirZhipu$0.1195free$0.023980%0
GLM-4.7-Flash (Deep Infra)Zhipu$0.06free$0.0183.3%0
Hy3 (Deep Infra)—$0.13free$0.03374.6%0
Kimi K2 Thinking (Amazon Bedrock)Amazon$0.60—$0.3050%—
Mistral Small (latest)Mistral$0.15—$0.01590%—
Qwen3.8 2.4T A95BAlibaba$2.00$2.50+25%$0.2587.5%1
GPT-5.4 ProOpenAI$30.00$30.00$30.000.0%never
Inkling (Deep Infra)Databricks$0.95$1.00+5.3%$0.1683.2%1
MiMo-V2.5 (Free)—freefreefree—never
MiMo-V2.5-Pro—$0.52freefree100%0
o3OpenAI$2.00free$0.5075%0
Qwen3.5 35B-A3BAlibaba$0.057—$0.0570.0%—
Step 3.7 Flash (Deep Infra)—$0.20—$0.0480%—
DeepSeek V3.2DeepSeek$3.00—$0.1395.7%—
GLM-4.5Zhipu$0.60free$0.1181.7%0
GPT-5.2 Codex (Azure)OpenAI$1.75—$0.17590%—
GPT-5.5 ProOpenAI$30.00$30.00$30.000.0%never
Qwen/Qwen3-32BAlibaba$0.1494—$0.0566.5%—
Qwen3.5 9BAlibaba$0.09$0.1875+108.3%$0.04550%3
Step 5 Preview (Deep Infra)—$1.00—$0.1090%—
GPT-4.1 nano (Azure)OpenAI$0.10free$0.02575%0
Hy4 preview (Deep Infra)—$0.834$0.96+15.1%$0.04295.0%1
Kimi K3.1 ProKimi$2.00$2.00$0.2090%0
Qwen3 235B-A22B Instruct 2507 (Nebius)Alibaba$0.20—$0.200.0%—
Qwen3.5 122B-A10BAlibaba$0.115$0.625+443.5%$0.1150.0%never
Qwen3.5 27BAlibaba$0.086—$0.0860.0%—
Qwen3.5 PlusAlibaba$0.36$0.40+11.1%$0.360.0%never
claude-opus-4-1-20250805Anthropic$15.00$18.75+25%$1.5090%1
GPT-5.1 Codex (Azure)OpenAI$1.25—$0.12590%—
GPT-5.1 Codex mini (Azure)OpenAI$0.25—$0.02590%—
Grok Build 0.1xAI$1.00$1.00$0.2080%0
MiniMax-M2MiniMax$0.30$0.375+25%$0.0390%1
Qwen3 Coder Next (IONOS)Amazon$0.1676free$0.0758.2%0
Qwen3 Max (Deep Infra)Alibaba$1.20free$0.2480%0
Qwen3 VL 235B A22B Instruct (Deep Infra)Alibaba$0.20—$0.1145%—
Codestral 25.08Mistral$0.30—$0.0390%—
DeepSeek V3 0324DeepSeek$0.275—$0.029989.1%—
DeepSeek V4 Flash Vision ExpDeepSeek$0.15$0.30+100%$0.00398%2
Gemma 3 27B IT (Nebius)Amazon$0.10—$0.100.0%—
GLM-4.5VZhipu$0.2988free$0.059880.0%0
GLM-5-TurboZhipu$1.20free$0.2480%0
MiniMax M2.7 HighspeedMiniMax$0.30$0.375+25%$0.0390%1
Qwen3.6 FlashAlibaba$0.25free$0.250.0%never
GLM-5V-TurboZhipu$1.20free$0.2480%0
Muse Spark 1.2 (Meta)Meta$1.25—$0.1588%—
o4-miniOpenAI$0.99free$0.2574.7%0
Qwen3-30B-A3BAlibaba$0.11—$0.0554.5%—
Qwen3-Coder 30B-A3B Instruct (Scaleway)Alibaba$0.2245—$0.01493.8%—
GLM-4.6VZhipu$0.30free$0.0583.3%0
GPT-5 ProOpenAI$15.00—$0.005100.0%—
Muse Spark 1.3 (Meta)Meta$1.25—$0.1588%—
Muse Spark Open (Nebius)Meta$1.25—$1.250.0%—
Qwen3-235B-A22BAlibaba$0.29$0.09$0.01893.8%0
GPT-5.1 Codex Max (Azure)OpenAI$1.25—$0.12590%—
Kimi K2.7 Code HighspeedKimi$1.90free$0.3880%0
Muse Glimmer 30B (Deep Infra)Meta$0.30—$0.0486.7%—
Muse Spark 1.1 (Meta)Meta$1.25—$0.1588%—
Nano Banana Pro (Vertex AI)Google$2.00$0.375$0.2090%0
o3-mini (Microsoft)Microsoft$1.10free$0.5550%0

150 of these models list a cache-read price and 120 list a cache-write price. Break-even is blank for the rest: without price_cache_write the write premium is unknown, and assuming it is zero would put a confident 0 in 30 rows nobody measured. There is no TTL column and no minimum-token column because no source we ingest publishes either field.

A worked example

One million prefix tokens on MiniMax-M3, sent once and reused 10 times. Cached costs $0.90 to write plus 10 reads at $0.06, so $1.50. Uncached costs 11 sends at $0.30, so $3.30. Break-even for this model is 3 later requests.

Common questions

What is prompt caching?
A host stores the tokens at the start of your prompt. A later request that repeats that prefix is charged a cache-read rate instead of the input rate. The saving is on input tokens only. Output is charged as usual.
How much does a cache hit save?
The discount column is (input price − cache-read price) ÷ input price, from the prices each host lists. It is not a measurement. Most hosts list a cache read at one tenth of their input price.
What is the break-even column?
The smallest number of later requests at which caching a prefix costs less than sending it uncached every time. Writing once and reading n times costs (cache write + n × cache read). Not caching costs (1 + n) × input. The column is blank when the host lists no cache-write price, because the write premium is then unknown.
Why is there no cache TTL or minimum-token column?
No source this site ingests publishes either one. Checked on 2026-09-02 against every corpus row: the fields available are the cache-read price, the cache-write price, a longer-TTL write price at a few hosts, and a caching capability flag. A TTL column would be invented, so there is none.
Do all hosts of one model charge the same cache price?
No. A price belongs to a model and a host. This table shows one listing per model. Open the model page to see every host.