API prices

Every tier the calculator knows about, on one page. Headline rates are what providers publish; the last column is what a request of your shape actually costs on each, which is the only figure that ranks them like for like. Researched to 2026-10-07.

Price them against your own request

Read 98% Written 2% Full rate 0%

98% of input tokens billed as cache hits, 2% as cache writes, 0% at the plain input rate. Measured, not assumed: the token-weighted mean of 64,680 recorded coding-agent requests. Almost every billed input token is a cache read, because a conversation re-reads its whole prefix every turn while writing only that turn’s delta — reads grow with the square of the turn count and writes only linearly. Output is 0.4% of input, so this is a prefill-bound workload, not a decode-bound one.

54 tiers · 29 superseded hidden · 3 too small for this request, sorted last

API price tiers with their per-million rates as actually billed — batch and flex discounts already applied — and the cost of one request at the selected workload shape.
TierScoreInputCached inCache writeOutputContextAt your mix
Xiaomi · MiMo V2.6 Flash Batch · 50% multiplier window too small for this requestinput limit too small for this requestoutput cap too small for this requestoutput cap may be too small rates shown after the 50% batch discount 38 of 100 on Artificial Analysis Intelligence Index v4.3.2, observed 2026-10-07 (opens in a new tab)$0.070$0.0014$0$0.1401.05M$0.0020 per M tokens
Xiaomi · MiMo V2.6 Pro Batch · 50% multiplier window too small for this requestinput limit too small for this requestoutput cap too small for this requestoutput cap may be too small rates shown after the 50% batch discount 46 of 100 on Artificial Analysis Intelligence Index v4.3.2, observed 2026-09-25 (opens in a new tab)$0.217$0.0018$0$0.4351.05M$0.0036 per M tokens 1.8×
Xiaomi · MiMo V2.6 Flash standard window too small for this requestinput limit too small for this requestoutput cap too small for this requestoutput cap may be too small 38 of 100 on Artificial Analysis Intelligence Index v4.3.2, observed 2026-10-07 (opens in a new tab)$0.140$0.0028$0$0.2801.05M$0.0039 per M tokens 2×
Meta · Muse Spark 1.2 contributor · Meta trains on your requests trains on requests window too small for this requestinput limit too small for this requestoutput cap too small for this requestoutput cap may be too small 40 of 100 on Artificial Analysis Intelligence Index v4.3.2, observed 2026-09-25 (opens in a new tab)$0.100$0.0020—$0.2001.05M$0.0048 per M tokens 2.4×
Meta · Muse Spark 1.3 contributor · Meta trains on your requests trains on requests window too small for this requestinput limit too small for this requestoutput cap too small for this requestoutput cap may be too small 48 of 100 on Artificial Analysis Intelligence Index v4.3.2, observed 2026-09-25 (opens in a new tab)$0.100$0.0020—$0.2001.05M$0.0048 per M tokens 2.4×
OpenAI · GPT-6 Luna Batch / Flex · 50% multiplier window too small for this requestinput limit too small for this requestoutput cap too small for this requestoutput cap may be too small rates shown after the 50% batch discount 37 of 100 on Artificial Analysis Intelligence Index v4.3.2, observed 2026-09-25 (opens in a new tab)$0.050$0.0050$0.063$0.2501.05M$0.0072 per M tokens 3.7×
Xiaomi · MiMo V2.6 Pro standard window too small for this requestinput limit too small for this requestoutput cap too small for this requestoutput cap may be too small 46 of 100 on Artificial Analysis Intelligence Index v4.3.2, observed 2026-09-25 (opens in a new tab)$0.435$0.0036$0$0.8701.05M$0.0072 per M tokens 3.7×
DeepSeek · DeepSeek V4.1 Flash peak / off-peak · from 2026-09-10 window too small for this requestinput limit too small for this requestoutput cap too small for this requestoutput cap may be too small 39 of 100 on Artificial Analysis Intelligence Index v4.3.2, observed 2026-09-25 (opens in a new tab)$0.300$0.0060$0.300$1.201.05M$0.010 per M tokens 5.2×
OpenAI · GPT-6 Luna standard window too small for this requestinput limit too small for this requestoutput cap too small for this requestoutput cap may be too small 37 of 100 on Artificial Analysis Intelligence Index v4.3.2, observed 2026-09-25 (opens in a new tab)$0.100$0.010$0.125$0.5001.05M$0.014 per M tokens 7.3×
Google · Gemini 3.1 Flash-Lite Batch / Flex · 50% multiplier window too small for this requestinput limit too small for this requestoutput cap too small for this requestoutput cap may be too small rates shown after the 50% batch discount—$0.125$0.013$0.500 per h stored—$0.7501.05M$0.018 per M tokens 9.1×
Alibaba · Qwen3.8-Flash standard window too small for this requestinput limit too small for this requestoutput cap too small for this requestoutput cap may be too small —$0.150$0.016$0.150$0.4701M$0.021 per M tokens 10.5×
Google · Gemini 3.5 Flash-Lite Batch / Flex window too small for this requestinput limit too small for this requestoutput cap too small for this requestoutput cap may be too small 22 of 100 on Artificial Analysis Intelligence Index v4.3.2, observed 2026-09-25 (opens in a new tab)$0.150$0.020$1.00 per h stored—$1.251.05M$0.028 per M tokens 14.2×
Z.ai · GLM-5.3-Flash list rate · from 2026-09-10 window too small for this requestinput limit too small for this requestoutput cap too small for this requestoutput cap may be too small 42 of 100 on Artificial Analysis Intelligence Index v4.3.2, observed 2026-09-25 (opens in a new tab)$0.150$0.030—$0.5001.05M$0.034 per M tokens 17.6×
Google · Gemini 3.1 Flash-Lite standard window too small for this requestinput limit too small for this requestoutput cap too small for this requestoutput cap may be too small —$0.250$0.025$1.00 per h stored—$1.501.05M$0.036 per M tokens 18.2×
Google · Gemini 3.5 Flash-Lite standard window too small for this requestinput limit too small for this requestoutput cap too small for this requestoutput cap may be too small 22 of 100 on Artificial Analysis Intelligence Index v4.3.2, observed 2026-09-25 (opens in a new tab)$0.300$0.030$1.00 per h stored—$2.501.05M$0.046 per M tokens 23.4×
DeepSeek · DeepSeek V4 Pro peak / off-peak · from 2026-08-16 window too small for this requestinput limit too small for this requestoutput cap too small for this requestoutput cap may be too small 36 of 100 on Artificial Analysis Intelligence Index v4.3.2, observed 2026-09-25 (opens in a new tab)$1.32$0.044$1.32$3.961.05M$0.052 per M tokens 26.5×
Google · Gemini 3.7 Flash Batch / Flex · 50% multiplier · promotional through 2026-12-31 window too small for this requestinput limit too small for this requestoutput cap too small for this requestoutput cap may be too small rates shown after the 50% batch discount 39 of 100 on Artificial Analysis Intelligence Index v4.3.2, observed 2026-09-25 (opens in a new tab)$0.375$0.037$0.500 per h stored—$1.881.05M$0.052 per M tokens 26.6×
Google · Gemini 3.8 Flash Batch / Flex · 50% multiplier · promotional through 2026-12-31 window too small for this requestinput limit too small for this requestoutput cap too small for this requestoutput cap may be too small rates shown after the 50% batch discount 41 of 100 on Artificial Analysis Intelligence Index v4.3.2, observed 2026-09-25 (opens in a new tab)$0.375$0.037$0.500 per h stored—$1.881.05M$0.052 per M tokens 26.6×
Google · Gemini 3.7 Flash standard · promotional through 2026-12-31 window too small for this requestinput limit too small for this requestoutput cap too small for this requestoutput cap may be too small 39 of 100 on Artificial Analysis Intelligence Index v4.3.2, observed 2026-09-25 (opens in a new tab)$0.750$0.075$0.500 per h stored—$3.751.05M$0.104 per M tokens 53.1×
Google · Gemini 3.7 Flash Batch / Flex · 50% multiplier · list rate from 2027-01-01 window too small for this requestinput limit too small for this requestoutput cap too small for this requestoutput cap may be too small rates shown after the 50% batch discount 39 of 100 on Artificial Analysis Intelligence Index v4.3.2, observed 2026-09-25 (opens in a new tab)$0.750$0.075$1.00 per h stored—$3.751.05M$0.104 per M tokens 53.1×
Google · Gemini 3.8 Flash standard · promotional through 2026-12-31 window too small for this requestinput limit too small for this requestoutput cap too small for this requestoutput cap may be too small 41 of 100 on Artificial Analysis Intelligence Index v4.3.2, observed 2026-09-25 (opens in a new tab)$0.750$0.075$0.500 per h stored—$3.751.05M$0.104 per M tokens 53.1×
Google · Gemini 3.8 Flash Batch / Flex · 50% multiplier · list rate from 2027-01-01 window too small for this requestinput limit too small for this requestoutput cap too small for this requestoutput cap may be too small rates shown after the 50% batch discount 41 of 100 on Artificial Analysis Intelligence Index v4.3.2, observed 2026-09-25 (opens in a new tab)$0.750$0.075$1.00 per h stored—$3.751.05M$0.104 per M tokens 53.1×
Anthropic · Claude Sonnet 5 Batch API estimated window too small for this requestinput limit too small for this requestoutput cap too small for this requestoutput cap may be too small rates shown after the 50% batch discount 38 of 100 on Artificial Analysis Intelligence Index v4.3.2, observed 2026-09-25 (opens in a new tab)$1.00$0.100$1.25$2.00 for 1 h$5.001M$0.144 per M tokens 73.4×
OpenAI · GPT-6 Sol Batch / Flex · 50% multiplier window too small for this requestinput limit too small for this requestoutput cap too small for this requestoutput cap may be too small rates shown after the 50% batch discount 48 of 100 on Artificial Analysis Intelligence Index v4.3.2, observed 2026-09-25 (opens in a new tab)$1.00$0.100$1.25$5.001.05M$0.144 per M tokens 73.4×
OpenAI · GPT-5.6 Terra Flex / Batch · 50% multiplier window too small for this requestinput limit too small for this requestoutput cap too small for this requestoutput cap may be too small rates shown after the 50% batch discount 42 of 100 on Artificial Analysis Intelligence Index v4.3.2, observed 2026-09-25 (opens in a new tab)$1.00$0.100$1.25$6.001.05M$0.148 per M tokens 75.5×
Meta · Muse Spark 1.2 standard window too small for this requestinput limit too small for this requestoutput cap too small for this requestoutput cap may be too small 40 of 100 on Artificial Analysis Intelligence Index v4.3.2, observed 2026-09-25 (opens in a new tab)$1.25$0.150—$4.251.05M$0.189 per M tokens 96.7×
Meta · Muse Spark 1.3 standard window too small for this requestinput limit too small for this requestoutput cap too small for this requestoutput cap may be too small 48 of 100 on Artificial Analysis Intelligence Index v4.3.2, observed 2026-09-25 (opens in a new tab)$1.25$0.150—$4.251.05M$0.189 per M tokens 96.7×
Anthropic · Claude Opus 5.5 Batch API estimated window too small for this requestinput limit too small for this requestoutput cap too small for this requestoutput cap may be too small rates shown after the 50% batch discount 58 of 100 on Artificial Analysis Intelligence Index v4.3.2, observed 2026-09-25 (opens in a new tab)$2.00$0.100$2.50$4.00 for 1 h$10.001M$0.190 per M tokens 96.9×
Google · Gemini 3.7 Flash list rate · from 2027-01-01 window too small for this requestinput limit too small for this requestoutput cap too small for this requestoutput cap may be too small 39 of 100 on Artificial Analysis Intelligence Index v4.3.2, observed 2026-09-25 (opens in a new tab)$1.50$0.150$1.00 per h stored—$7.501.05M$0.208 per M tokens 106.3×
Google · Gemini 3.8 Flash list rate · from 2027-01-01 window too small for this requestinput limit too small for this requestoutput cap too small for this requestoutput cap may be too small 41 of 100 on Artificial Analysis Intelligence Index v4.3.2, observed 2026-09-25 (opens in a new tab)$1.50$0.150$1.00 per h stored—$7.501.05M$0.208 per M tokens 106.3×
Alibaba · Qwen3.7-Plus standard window too small for this requestinput limit too small for this requestoutput cap too small for this requestoutput cap may be too small 25 of 100 on Artificial Analysis Intelligence Index v4.3.2, observed 2026-09-25 (opens in a new tab)$0.400$0.080$0.400$1.601M$0.278 per M tokens 142.3×
Anthropic · Claude Sonnet 5 standard window too small for this requestinput limit too small for this requestoutput cap too small for this requestoutput cap may be too small 38 of 100 on Artificial Analysis Intelligence Index v4.3.2, observed 2026-09-25 (opens in a new tab)$2.00$0.200$2.50$4.00 for 1 h$10.001M$0.287 per M tokens 146.8×
OpenAI · GPT-6 Sol standard window too small for this requestinput limit too small for this requestoutput cap too small for this requestoutput cap may be too small 48 of 100 on Artificial Analysis Intelligence Index v4.3.2, observed 2026-09-25 (opens in a new tab)$2.00$0.200$2.50$10.001.05M$0.287 per M tokens 146.8×
OpenAI · GPT-5.6 Terra standard window too small for this requestinput limit too small for this requestoutput cap too small for this requestoutput cap may be too small 42 of 100 on Artificial Analysis Intelligence Index v4.3.2, observed 2026-09-25 (opens in a new tab)$2.00$0.200$2.50$12.001.05M$0.296 per M tokens 151.1×
Z.ai · GLM-5.3 standard window too small for this requestinput limit too small for this requestoutput cap too small for this requestoutput cap may be too small 45 of 100 on Artificial Analysis Intelligence Index v4.3.2, observed 2026-09-25 (opens in a new tab)$1.40$0.260—$4.401.05M$0.300 per M tokens 153.4×
Z.ai · GLM-5.2 standard window too small for this requestinput limit too small for this requestoutput cap too small for this requestoutput cap may be too small 34 of 100 on Artificial Analysis Intelligence Index v4.3.2, observed 2026-09-25 (opens in a new tab)$1.40$0.260—$4.401.05M$0.300 per M tokens 153.4×
Alibaba · Qwen3.8-Max standard · snapshot 0902 · international window too small for this requestinput limit too small for this requestoutput cap too small for this requestoutput cap may be too small 45 of 100 on Artificial Analysis Intelligence Index v4.3.2, observed 2026-09-25 (opens in a new tab)$2.00$0.250$2.00$6.001M$0.309 per M tokens 158×
Anthropic · Claude Fable 5.1 Batch API estimated window too small for this requestinput limit too small for this requestoutput cap too small for this requestoutput cap may be too small rates shown after the 50% batch discount 53 of 100 on Artificial Analysis Intelligence Index v4.3.2, observed 2026-09-25 (opens in a new tab)$5.00$0.125$6.25$10.00 for 1 h$25.001M$0.352 per M tokens 179.8×
Anthropic · Claude Mythos 5.1 Batch API preview estimated window too small for this requestinput limit too small for this requestoutput cap too small for this requestoutput cap may be too small rates shown after the 50% batch discount—$5.00$0.125$6.25$10.00 for 1 h$25.001M$0.352 per M tokens 179.8×
Anthropic · Claude Opus 5.5 standard window too small for this requestinput limit too small for this requestoutput cap too small for this requestoutput cap may be too small 58 of 100 on Artificial Analysis Intelligence Index v4.3.2, observed 2026-09-25 (opens in a new tab)$4.00$0.200$5.00$8.00 for 1 h$20.001M$0.379 per M tokens 193.8×
Moonshot AI · Kimi K3 standard window too small for this requestinput limit too small for this requestoutput cap too small for this requestoutput cap may be too small 44 of 100 on Artificial Analysis Intelligence Index v4.3.2, observed 2026-09-25 (opens in a new tab)$3.00$0.300$3.00$6.00 for 1 h$15.001.05M$0.416 per M tokens 212.5×
Google · Gemini 3.1 Pro Preview · Batch / Flex · 50% multiplier · ≤200K preview window too small for this requestinput limit too small for this requestoutput cap too small for this requestoutput cap may be too small rates shown after the 50% batch discount 30 of 100 on Artificial Analysis Intelligence Index v4.3.2, observed 2026-09-25 (opens in a new tab)$1.00$0.200$4.50 per h stored—$6.001.05M$0.468 per M tokens 239.3×
Google · Gemini 3.1 Pro Preview · ≤200K preview window too small for this requestinput limit too small for this requestoutput cap too small for this requestoutput cap may be too small 30 of 100 on Artificial Analysis Intelligence Index v4.3.2, observed 2026-09-25 (opens in a new tab)$2.00$0.200$4.50 per h stored—$12.001.05M$0.546 per M tokens 279.1×
Alibaba · Qwen3.7-Max standard window too small for this requestinput limit too small for this requestoutput cap too small for this requestoutput cap may be too small 29 of 100 on Artificial Analysis Intelligence Index v4.3.2, observed 2026-09-25 (opens in a new tab)$2.50$0.500$2.50$7.501M$0.569 per M tokens 291.1×
Anthropic · Claude Fable 5.1 standard window too small for this requestinput limit too small for this requestoutput cap too small for this requestoutput cap may be too small 53 of 100 on Artificial Analysis Intelligence Index v4.3.2, observed 2026-09-25 (opens in a new tab)$10.00$0.250$12.50$20.00 for 1 h$50.001M$0.704 per M tokens 359.7×
Anthropic · Claude Mythos 5.1 limited availability preview window too small for this requestinput limit too small for this requestoutput cap too small for this requestoutput cap may be too small —$10.00$0.250$12.50$20.00 for 1 h$50.001M$0.704 per M tokens 359.7×
OpenAI · GPT-6 Astra Batch / Flex · 50% multiplier window too small for this requestinput limit too small for this requestoutput cap too small for this requestoutput cap may be too small rates shown after the 50% batch discount 53 of 100 on Artificial Analysis Intelligence Index v4.3.2, observed 2026-09-25 (opens in a new tab)$5.00$0.500$6.25$25.001.05M$0.718 per M tokens 366.9×
Anthropic · Claude Opus 5.5 fast mode preview estimated window too small for this requestinput limit too small for this requestoutput cap too small for this requestoutput cap may be too small —$8.00$0.400$10.00$16.00 for 1 h$40.001M$0.758 per M tokens 387.5×
xAI · Grok 4.7 standard · auto long-context ≥200K window too small for this requestinput limit too small for this requestoutput cap too small for this requestoutput cap may be too small 46 of 100 on Artificial Analysis Intelligence Index v4.3.2, observed 2026-09-25 (opens in a new tab)$2.00$0.500—$6.00500K$1.11 per M tokens 565.5×
OpenAI · GPT-6 Astra standard window too small for this requestinput limit too small for this requestoutput cap too small for this requestoutput cap may be too small 53 of 100 on Artificial Analysis Intelligence Index v4.3.2, observed 2026-09-25 (opens in a new tab)$10.00$1.00$12.50$50.001.05M$1.44 per M tokens 733.9×
OpenAI · GPT-5.6 Cyber standard window too small for this requestinput limit too small for this requestoutput cap too small for this requestoutput cap may be too small —$12.50$1.25$15.63$75.00400K$1.85 per M tokens 944.3×
Mistral AI · Mistral Medium 3.5 Batch · 50% multiplier estimated window too small for this requestinput limit too small for this requestoutput cap too small for this requestoutput cap may be too small rates shown after the 50% batch discount 14 of 100 on Artificial Analysis Intelligence Index v4.3.2, observed 2026-09-25 (opens in a new tab)$0.750$0.075to $0.150—$3.75256K$0.104 per M tokens
Anthropic · Claude Haiku 4.5 standard window too small for this requestinput limit too small for this requestoutput cap too small for this requestoutput cap may be too small 17 of 100 on Artificial Analysis Intelligence Index v4.3.2, observed 2026-09-26 (opens in a new tab)$1.00$0.100$1.25$2.00 for 1 h$5.00200K$0.144 per M tokens
Mistral AI · Mistral Medium 3.5 standard window too small for this requestinput limit too small for this requestoutput cap too small for this requestoutput cap may be too small 14 of 100 on Artificial Analysis Intelligence Index v4.3.2, observed 2026-09-25 (opens in a new tab)$1.50$0.150—$7.50256K$0.208 per M tokens

How to read this

  • The first six columns are the provider's published rates, per million tokens, except in a tier marked estimated, where at least one of them is derived here from other published figures and the tier's note in the catalog says which. A dash means the provider does not publish that line at all — not that it is free. Where no cache-hit rate exists, the calculator charges the ordinary input rate. Cache write is the provider's default entry. Where it also sells one that lives an hour, as Anthropic and Moonshot do beside a five-minute default, that price sits beneath it, and "At your mix" charges every write the default: the calculator's One-hour writes control is where a split is priced. A cached-input rate that runs "to" a second one is a provider that states its batch discount and its cache price but not whether the one reaches the other, as Mistral does: the first figure has the discount on top and is what "At your mix" charges, the second does not. A rate "per h stored" is what Google charges to keep a million cached tokens for an hour, on top of each hit. "At your mix" holds nothing and charges none of it, because how long a cache is kept depends on the reader's traffic: the calculator's Cache held per request control prices it.
  • "At your mix" is the only column that ranks tiers honestly. It prices one request of the shape set above on every tier and converts to dollars per million tokens, so caching, long-context surcharges, batch discounts and time-of-day windows are already inside the number. Change the shape and the order changes with it — which is the point.
  • A capability score beside a price is there to stop a category error. The cheapest row is frequently the cheapest because it answers an easier class of question. Sort by capability per dollar to see the trade rather than one half of it. A ≈ before a score means Artificial Analysis marks it as an estimate rather than a measurement.
  • Tiers with a discount window are priced at the share a continuous service would naturally land there — the share of the week the peak schedule leaves over, not the best case. Days count as well as hours: a peak that runs on weekdays only leaves both weekend days off-peak. A batch workload that can wait for the cheap hours does better; an interactive one does not.
    DeepSeek V4 Flash: peak 01:00–04:00 and 06:00–10:00 UTC, Mon–Fri, so 79% of a round-the-clock workload is costed off-peak.
    DeepSeek V4 Pro: peak 01:00–04:00 and 06:00–10:00 UTC, Mon–Fri, so 79% of a round-the-clock workload is costed off-peak.
    DeepSeek V4.1 Flash: peak 01:00–04:00 and 06:00–10:00 UTC, Mon–Fri, so 79% of a round-the-clock workload is costed off-peak.
  • Retired tiers stay resolvable but are hidden by default, so a shared scenario URL from three months ago still opens. Tick the box to see what a rate replaced.
Inference Economics · static, local-first calculationCatalog 0.11.0 · cutoff 2026-10-07 · app v0.11.0 · build 17df6ef