pricing table 2026-10-26 · verified against the providers' pages · USD per 1M tokens

LLM API prices, caching rates and cheaper siblings

The same table the gateway signs receipts with. Each model page shows the prompt-caching break-even and what routing to its cheaper sibling saves.

ModelProviderTierInputCached inputCache writeOutputCheaper siblingLifecycle
claude-fable-5-1anthropicfrontier$10$0.25$12.5$50claude-opus-5-5active
claude-fable-5anthropicfrontier$10$1$12.5$50claude-opus-5-5active
claude-opus-5anthropicfrontier$5$0.5$6.25$25claude-sonnet-5-5active
claude-opus-4-8anthropicfrontier$5$0.5$6.25$25claude-sonnet-5-5active
claude-opus-4-7anthropicfrontier$5$0.5$6.25$25claude-sonnet-5-5active
claude-opus-4-6anthropicfrontier$5$0.5$6.25$25claude-sonnet-4-6active
claude-opus-4-5anthropicfrontier$5$0.5$6.25$25—active
claude-opus-5-5anthropicfrontier$4$0.2$5$20claude-sonnet-5-5active
claude-sonnet-4-6anthropicmid$3$0.3$3.75$15claude-haiku-4-5active
claude-sonnet-4-5anthropicmid$3$0.3$3.75$15claude-haiku-4-5retires 2026-11-30 → claude-sonnet-5-5
claude-sonnet-5-5anthropicmid$2$0.1$2.5$10claude-haiku-5-5active
claude-sonnet-5anthropicmid$2$0.2$2.5$10claude-haiku-5-5active
claude-haiku-4-5anthropicfast$1$0.1$1.25$5—active
claude-haiku-5-5anthropicfast$0.1$0.01$0.125$0.5—active
deepseek-v4-prodeepseekmid$1.32$0.044—$3.96deepseek-flashactive
deepseek-flashdeepseekfast$0.3$0.006—$1.2—active
gemini-4-argongeminifrontier$2$0.1—$10gemini-3.8-flashactive
gemini-3.1-progeminifrontier$2$0.2—$12gemini-3.8-flashactive
gemini-2.5-progeminifrontier$1.25$0.125—$10gemini-2.5-flashactive
gemini-3.8-flashgeminifast$0.75$0.075—$3.75gemini-3.5-flash-liteactive
gemini-3.6-flashgeminifast$0.75$0.075—$3.75—active
gemini-3-flashgeminifast$0.5$0.05—$3—active
gemini-3.5-flash-litegeminifast$0.3$0.03—$2.5—active
gemini-2.5-flashgeminifast$0.3$0.03—$2.5gemini-2.5-flash-liteactive
gemini-3.1-flash-litegeminifast$0.25$0.025—$1.5—active
gemini-2.5-flash-litegeminifast$0.1$0.01—$0.4—active
gpt-5.5-proopenaifrontier$30——$180—active
gpt-5.4-proopenaifrontier$30——$180—active
gpt-5.2-proopenaifrontier$21——$168—active
o3-proopenaifrontier$20——$80—retires 2026-12-11 → gpt-5.6-sol
gpt-5-proopenaifrontier$15——$120—retires 2026-12-11 → gpt-5.6-sol
o1openaifrontier$15$7.5—$60—retired 2026-10-23 → gpt-5.6-sol
gpt-6-astraopenaifrontier$10$1$12.5$50gpt-6.1-solactive
gpt-5.5openaifrontier$5$0.5—$30gpt-5.4-miniactive
gpt-5.6-solopenaifrontier$4$0.4$5$20gpt-5.6-terraactive
gpt-5.4openaifrontier$2.5$0.25—$15gpt-5.4-miniactive
gpt-4oopenaimid$2.5$1.25—$10gpt-4o-miniretires 2027-01-14 → gpt-6-luna
gpt-6.1-solopenaimid$2$0.1$2.5$10gpt-6-lunaactive
gpt-6-solopenaimid$2$0.2$2.5$10gpt-6-lunaactive
gpt-5.6-terraopenaimid$2$0.2$2.5$12gpt-5.6-lunaactive
o3openaimid$2$0.5—$8o4-miniretires 2026-12-11 → gpt-5.6-sol
gpt-4.1openaimid$2$0.5—$8gpt-4.1-miniretires 2027-01-14 → gpt-6-luna
gpt-5.2openaifrontier$1.75$0.175—$14gpt-5-miniactive
gpt-5.1openaimid$1.25$0.125—$10gpt-5-miniretires 2027-04-01 → gpt-6-sol
gpt-5openaimid$1.25$0.125—$10gpt-5-miniretires 2026-12-11 → gpt-5.6-sol
o3-miniopenaifast$1.1$0.55—$4.4—retired 2026-10-23 → gpt-5.6-sol
o4-miniopenaifast$1.1$0.275—$4.4—retired 2026-10-23 → gpt-5.6-terra
gpt-5.4-miniopenaifast$0.75$0.075—$4.5—active
gpt-4.1-miniopenaifast$0.4$0.1—$1.6—retires 2027-01-14 → gpt-6-luna
gpt-5-miniopenaifast$0.25$0.025—$2—retires 2026-12-11 → gpt-5.6-terra
gpt-5.6-lunaopenaifast$0.2$0.02$0.25$1.2—active
gpt-5.4-nanoopenaifast$0.2$0.02—$1.25—retires 2027-04-01 → gpt-6-luna
gpt-4o-miniopenaifast$0.15$0.075—$0.6—retires 2027-01-14 → gpt-6-luna
gpt-6-lunaopenaifast$0.1$0.01$0.125$0.5—active
gpt-4.1-nanoopenaifast$0.1$0.025—$0.4—retired 2026-10-23 → gpt-5.6-luna
gpt-5-nanoopenaifast$0.05$0.005—$0.4—retires 2026-12-11 → gpt-5.6-luna
grok-4.7xaifrontier$2$0.5—$6grok-4.3active
grok-4.6xaifrontier$2$0.5—$6grok-4.3active
grok-4.5xaifrontier$2$0.3—$6grok-4.3active
grok-4.3xaimid$1.25$0.2—$2.5—active
grok-4.20xaimid$1.25$0.2—$2.5—active
grok-build-0.1xaifast$1$0.2—$2—active

Processing-tier multipliers

Applied to every token price; TokenZero reads the tier the provider served and prices the receipt with it.

TierMultiplier
standard1×
flex / batch (OpenAI)0.5×
fast (OpenAI; 2.5× on gpt-5.5, 1.7–1.82× on GPT-4.x and o-series; Anthropic Opus speed:fast 2×)2×
ultrafast (OpenAI, gpt-6-astra and gpt-6.1-sol only)6×
US-only inference (Anthropic inference_geo: us, Claude 4.6+)1.1×
off-peak (DeepSeek, outside Mon–Fri 01–04 / 06–10 UTC)0.5×