AI Pricing

Context is 128,000 input tokens; No cache-write unit price is listed.

• StandardInput USD 0.15 · Output USD 0.6 · Cache read USD 0.075

Context is 400,000 input tokens; No cache-write unit price is listed.

• StandardInput USD 1.25 · Output USD 10 · Cache read USD 0.125

Context is 400,000 input tokens; No cache-write unit price is listed.

• StandardInput USD 0.25 · Output USD 2 · Cache read USD 0.025

Context is 400,000 input tokens; No cache-write unit price is listed.

• StandardInput USD 0.05 · Output USD 0.4 · Cache read USD 0.005

The official GPT-5.3-Codex page lists a 400K context and 128K maximum output; it lists cached input but no cache-write price.

• StandardInput USD 1.75 · Output USD 14 · Cache read USD 0.175

Context is 1,050,000 input tokens; prompts over 272K use 2x input/cache and 1.5x output for the full request; regional processing adds 10%; No cache-write unit price is listed.

• StandardInput USD 2.5 · Output USD 15 · Cache read USD 0.25
• Long context (>272K, entire request)Input USD 5 · Output USD 22.5 · Cache read USD 0.5

Context is 400,000 input tokens; regional processing adds 10%; No cache-write unit price is listed.

• StandardInput USD 0.75 · Output USD 4.5 · Cache read USD 0.075

Context is 400,000 input tokens; regional processing adds 10%; No cache-write unit price is listed.

• StandardInput USD 0.2 · Output USD 1.25 · Cache read USD 0.02

Context is 1,050,000 input tokens; prompts over 272K use 2x input/cache and 1.5x output for the full request; regional processing adds 10%; No cache-write unit price is listed.

• StandardInput USD 5 · Output USD 30 · Cache read USD 0.5
• Long context (>272K, entire request)Input USD 10 · Output USD 45 · Cache read USD 1

Context is 1,050,000 input tokens; regional processing adds 10%; Responses API and Batch are supported; No cache-write unit price is listed. No cached-input discount is offered.

• StandardInput USD 30 · Output USD 180

Context is 1,050,000 input tokens; prompts over 272K use 2x input/cache and 1.5x output for the full request; regional processing adds 10%; Cache writes are 1.25x input ($0.25/M Standard).

• StandardInput USD 0.2 · Output USD 1.2 · Cache read USD 0.02
• Long context (>272K, entire request)Input USD 0.4 · Output USD 1.8 · Cache read USD 0.04

The official model is GPT-5.6 Sol; promotional pricing lasts at least through 2026-11-21. Inputs over 272K use long-context rates for the entire request; regional adds 10%, Batch/Flex are half price, Fast is 2x, and cache writes are 1.25x.

• Short context (≤272K)Input USD 4 · Output USD 20 · Cache read USD 0.4
• Long context (>272K, entire request)Input USD 8 · Output USD 30 · Cache read USD 0.8

Context is 1,050,000 input tokens; prompts over 272K use 2x input/cache and 1.5x output for the full request; regional processing adds 10%; Cache writes are 1.25x input ($2.5/M Standard).

• StandardInput USD 2 · Output USD 12 · Cache read USD 0.2
• Long context (>272K, entire request)Input USD 4 · Output USD 18 · Cache read USD 0.4

Inputs over 272K use long-context pricing for the entire request; regional endpoints add 10%, Batch/Flex are 50% of Standard, Fast is 2x, and cache writes are 1.25x input.

• Short context (≤272K)Input USD 10 · Output USD 50 · Cache read USD 1
• Long context (>272K, entire request)Input USD 20 · Output USD 75 · Cache read USD 2

Inputs over 272K use long-context pricing for the entire request; regional endpoints add 10%, Batch/Flex are 50% of Standard, Fast is 2x, and cache writes are 1.25x input.

• Short context (≤272K)Input USD 0.1 · Output USD 0.5 · Cache read USD 0.01
• Long context (>272K, entire request)Input USD 0.2 · Output USD 0.75 · Cache read USD 0.02

Context 1,050,000; over 272K the full request uses long-context rates, regional endpoints add 10%, Batch/Flex are half price, Fast is 2x.

• Short context (≤272K)Input USD 2 · Output USD 10 · Cache read USD 0.2
• Long context (>272K, entire request)Input USD 4 · Output USD 15 · Cache read USD 0.4

The model page lists a 1,050,000-token context. If input exceeds 272K, the entire request uses long-context rates; regional endpoints add 10%, Batch/Flex are 50% of Standard, Fast is 2x, and cache writes are 1.25x input.

• Short context (≤272K)Input USD 2 · Output USD 10 · Cache read USD 0.1
• Long context (>272K, entire request)Input USD 4 · Output USD 15 · Cache read USD 0.2

Context is 200,000 input tokens; the Batch toggle is $1 input/$0.25 cached/$4 output; No cache-write unit price is listed.

• StandardInput USD 2 · Output USD 8 · Cache read USD 0.5
• BatchInput USD 1 · Output USD 4 · Cache read USD 0.25

Context is 200,000 input tokens; Responses API and Batch are supported; No cache-write unit price is listed. No cached-input discount is offered.

• StandardInput USD 20 · Output USD 80

Five-minute writes are $12.50/M, one-hour writes $20/M, cache hits $1/M; Batch halves input/output and data-residency endpoints add 1.1x.

• StandardInput USD 10 · Output USD 50 · Cache read USD 1

Five-minute cache writes are $12.50/M, one-hour writes $20/M, and cache hits $0.25/M; first-party data-residency endpoints for Claude 4.6+ add 1.1x to all categories, and Batch halves input/output.

• StandardInput USD 10 · Output USD 50 · Cache read USD 0.25

Context is 200K with 64K maximum output; five-minute writes are $1.25/M, cache hits $0.10/M, and Batch halves input/output.

• StandardInput USD 1 · Output USD 5 · Cache read USD 0.1

Five-minute writes are $6.25/M, one-hour writes $10/M, cache hits $0.50/M; full 1M context is standard-priced, Batch halves input/output, and data-residency endpoints add 1.1x.

• StandardInput USD 5 · Output USD 25 · Cache read USD 0.5

First-party data-residency endpoints for Claude 4.6+ add 1.1x to all categories; Batch halves input/output.

• StandardInput USD 5 · Output USD 25 · Cache read USD 0.5

Five-minute writes are $6.25/M, one-hour writes $10/M, cache hits $0.50/M; Batch halves input/output and data-residency endpoints add 1.1x.

• StandardInput USD 5 · Output USD 25 · Cache read USD 0.5

Five-minute writes are $5/M, one-hour writes $8/M, cache hits $0.20/M; Fast is $8/$40, Batch halves input/output, and data-residency endpoints add 1.1x.

• StandardInput USD 4 · Output USD 20 · Cache read USD 0.2

Five-minute writes are $3.75/M, one-hour writes $6/M, cache hits $0.30/M; full 1M context is standard-priced, Batch halves input/output, and data-residency endpoints add 1.1x.

• StandardInput USD 3 · Output USD 15 · Cache read USD 0.3

Five-minute writes are $2.50/M, one-hour writes $4/M, cache hits $0.20/M; Batch halves input/output and data-residency endpoints add 1.1x.

• StandardInput USD 2 · Output USD 10 · Cache read USD 0.2

Five-minute writes are $2.50/M, one-hour writes $4/M, cache hits $0.10/M; the full 1M context uses standard rates, Batch halves input/output, and data-residency endpoints add 1.1x.

• StandardInput USD 2 · Output USD 10 · Cache read USD 0.1

Peak is Mon–Fri 01:00–04:00 and 06:00–10:00 UTC, excluding Chinese public holidays; Beijing time is 09:00–12:00 and 14:00–18:00 on working days. Off-peak is half rate. The catalog deepseek-v4 is served by DeepSeek-V4-Pro-0813.

• Peak: weekdays UTC 01–04 and 06–10 (excluding Chinese public holidays)Input USD 1.32 · Output USD 3.96 · Cache read USD 0.044
• Off-peak: all other timesInput USD 0.66 · Output USD 1.98 · Cache read USD 0.022

Peak is Mon–Fri 01:00–04:00 and 06:00–10:00 UTC, excluding Chinese public holidays; in Beijing time this is 09:00–12:00 and 14:00–18:00 on working days. All other times, including weekends and holidays, are off-peak at half rates. The retired deepseek-v4-flash alias is still accepted and routes to DeepSeek-V4.1-Flash-1017.

• Peak: weekdays UTC 01–04 and 06–10 (excluding Chinese public holidays)Input USD 0.3 · Output USD 1.2 · Cache read USD 0.006
• Off-peak: all other timesInput USD 0.15 · Output USD 0.6 · Cache read USD 0.003

The catalog display name is Jamba Large 1.7. AI21's official Jamba page lists version 1.7 (snapshot 2025-07), a 256K maximum token window, and API endpoint `jamba-large`, currently pointing to `jamba-large-1.7-2025-07`. The retired 1.5 snapshot is not the current 1.7 model; no cached-input price is published on the official pages.

• StandardInput USD 2 · Output USD 8

The catalog display name is Jamba Mini 2. AI21's official Jamba page lists version 2 (snapshot 2026-01), a 256K maximum token window, and API endpoint `jamba-mini`, currently pointing to `jamba-mini-2-2026-01`. The retired 1.5 snapshot is not the current Mini 2 model; no cached-input price is published on the official pages.

• StandardInput USD 0.2 · Output USD 0.4

Official rates are ¥12/¥36 for global/China and ¥18.736/¥56.207 for US and international; Batch halves input/output. The alias equals qwen3.7-max-2026-05-20; cache is discounted but no fixed hit rate is published in the table.

• Global/China (≤1M)Input CNY 12 · Output CNY 36
• US (≤1M)Input CNY 18.736 · Output CNY 56.207
• International (≤1M)Input CNY 18.736 · Output CNY 56.207
• Batch global/ChinaInput CNY 6 · Output CNY 18

Global/China list rates are ¥2/¥8 up to 256K and ¥6/¥24 for 256K–1M, currently 20% off at ¥1.6/¥6.4 and ¥4.8/¥19.2. US list rates ¥2.998/¥11.991 and ¥8.993/¥35.972 receive the same discount. Batch halves input/output; no promotion end date or fixed cache rate is published.

• Global/China ≤256K (20% promotion)Input CNY 1.6 · Output CNY 6.4
• Global/China 256K–1M (20% promotion)Input CNY 4.8 · Output CNY 19.2
• US ≤256K (20% promotion)Input CNY 2.3984 · Output CNY 9.5928
• US 256K–1M (20% promotion)Input CNY 7.1944 · Output CNY 28.7776

Official CNY rates are ¥0.8/¥2.7 for global/China and ¥1.094/¥3.427 for US/international; cache is supported but qwen3.8-flash cache-hit pricing is a console exception without a fixed published value.

• Global/China (≤1M)Input CNY 0.8 · Output CNY 2.7
• US/International (≤1M)Input CNY 1.094 · Output CNY 3.427

Official CNY rates are ¥12/¥36 for global/China and ¥14.988/¥44.965 for US and international; Batch halves input/output. Context cache is discounted, but qwen3.8-max cache-hit pricing is a console exception without a fixed published value.

• Global/China (≤1M)Input CNY 12 · Output CNY 36
• US (≤1M)Input CNY 14.988 · Output CNY 44.965
• International (≤1M)Input CNY 14.988 · Output CNY 44.965
• Batch global/ChinaInput CNY 6 · Output CNY 18

Amazon Bedrock Standard is $0.30/$2.50 per MTok; cache reads are 75% below on-demand input ($0.075/M). Priority, Flex, and Batch are $0.525/$4.375, $0.15/$1.25, and $0.15/$1.25.

• StandardInput USD 0.3 · Output USD 2.5 · Cache read USD 0.075
• PriorityInput USD 0.525 · Output USD 4.375 · Cache read USD 0.075
• FlexInput USD 0.15 · Output USD 1.25 · Cache read USD 0.075
• BatchInput USD 0.15 · Output USD 1.25 · Cache read USD 0.075

Amazon Nova Lite has 300K context; Standard is $0.06/$0.24 per MTok, cache reads are 75% below on-demand input ($0.015/M), and Batch is $0.03/$0.12.

• StandardInput USD 0.06 · Output USD 0.24 · Cache read USD 0.015
• BatchInput USD 0.03 · Output USD 0.12 · Cache read USD 0.015

Amazon Nova Micro has 128K context; Standard is $0.035/$0.14 per MTok, cache reads are 75% below on-demand input ($0.00875/M), and Batch is $0.0175/$0.07.

• StandardInput USD 0.035 · Output USD 0.14 · Cache read USD 0.00875
• BatchInput USD 0.0175 · Output USD 0.07 · Cache read USD 0.00875

Amazon Nova Pro has 300K context; Standard is $0.80/$3.20 per MTok, cache reads are 75% below on-demand input ($0.20/M), Priority is $1.40/$5.60, and Flex/Batch are $0.40/$1.60.

• StandardInput USD 0.8 · Output USD 3.2 · Cache read USD 0.2
• PriorityInput USD 1.4 · Output USD 5.6 · Cache read USD 0.2
• FlexInput USD 0.4 · Output USD 1.6 · Cache read USD 0.2
• BatchInput USD 0.4 · Output USD 1.6 · Cache read USD 0.2

The exact model page and official model list were checked; context is 128,000 and maximum output 64,000. Trial and production keys are free until rate limits are reached; over-limit input/output/cache prices are unpublished, so fields remain null.

Cohere's official FAQ lists Command R 03-2024 at $0.50 input and $1.50 output per MTok; it lists no cache price or context for this legacy model.

• StandardInput USD 0.5 · Output USD 1.5

Cohere's official FAQ lists Command R+ 08-2024 at $2.50 input and $10 output per MTok; it lists no cache price or context for this legacy model.

• StandardInput USD 2.5 · Output USD 10

Standard input/output/cached prices are $0.30/$2.50/$0.03 per 1M tokens; Batch/Flex/Priority have separate rates.

• StandardInput USD 0.3 · Output USD 2.5 · Cache read USD 0.03

Standard input/output/cached prices are $0.10/$0.40/$0.01 per 1M tokens; Batch/Flex/Priority have separate rates.

• StandardInput USD 0.1 · Output USD 0.4 · Cache read USD 0.01

≤200K input is $1.25/$10/$0.125; requests over 200K use $2.50/$15/$0.25 for the whole request; Batch/Flex are discounted.

• ≤200K inputInput USD 1.25 · Output USD 10 · Cache read USD 0.125
• >200K inputInput USD 2.5 · Output USD 15 · Cache read USD 0.25

Standard input/output/cached prices are $0.25/$1.50/$0.025 per 1M tokens; Batch/Flex/Priority have separate rates.

• StandardInput USD 0.25 · Output USD 1.5 · Cache read USD 0.025

The official ID is gemini-3.1-pro-preview; ≤200K is $2/$12/$0.20 and requests over 200K use $4/$18/$0.40 for the whole request; Batch/Flex are discounted.

• ≤200K inputInput USD 2 · Output USD 12 · Cache read USD 0.2
• >200K inputInput USD 4 · Output USD 18 · Cache read USD 0.4

The page lists 1,048,576 input and 65,536 output tokens and auto-routes to gemini-3.6-flash. Routed rates are $0.75/$3.75/$0.075 through 2026-12-31 and $1.50/$7.50/$0.15 from 2027-01-01; Batch/Flex have separate rates.

• Routed rate through 2026-12-31Input USD 0.75 · Output USD 3.75 · Cache read USD 0.075
• Routed rate from 2027-01-01Input USD 1.5 · Output USD 7.5 · Cache read USD 0.15

Standard input/output/cached prices are $0.30/$2.50/$0.03 per 1M tokens; Batch/Flex/Priority/Contributor have separate rates.

• StandardInput USD 0.3 · Output USD 2.5 · Cache read USD 0.03

The page states the current rate through 2026-12-31; from 2027-01-01 input/output/cache hit become $1.50/$7.50/$0.15; Batch/Flex have separate discounts.

• Introductory through 2026-12-31Input USD 0.75 · Output USD 3.75 · Cache read USD 0.075
• From 2027-01-01Input USD 1.5 · Output USD 7.5 · Cache read USD 0.15

The page lists 1,048,576 input and 65,536 output tokens and auto-routes to gemini-3.8-flash. Routed rates are $0.75/$3.75/$0.075 through 2026-12-31 and $1.50/$7.50/$0.15 from 2027-01-01; Batch/Flex have separate rates.

• Routed rate through 2026-12-31Input USD 0.75 · Output USD 3.75 · Cache read USD 0.075
• Routed rate from 2027-01-01Input USD 1.5 · Output USD 7.5 · Cache read USD 0.15

The page states the current rate through 2026-12-31; from 2027-01-01 input/output/cache hit become $1.50/$7.50/$0.15. Cache storage is $0.50/hour now and $1/hour afterward; Batch/Flex have separate discounts.

• Introductory through 2026-12-31Input USD 0.75 · Output USD 3.75 · Cache read USD 0.075
• From 2027-01-01Input USD 1.5 · Output USD 7.5 · Cache read USD 0.15

M3 has a 1M context. Standard ≤512K is $0.30/$1.20/$0.06, >512K is $0.60/$2.40/$0.12; Priority is 1.5x Standard. >512K currently has limited access and cache-write pricing is not listed.

• Standard ≤512KInput USD 0.3 · Output USD 1.2 · Cache read USD 0.06
• Priority ≤512KInput USD 0.45 · Output USD 1.8 · Cache read USD 0.09
• Standard >512KInput USD 0.6 · Output USD 2.4 · Cache read USD 0.12
• Priority >512KInput USD 0.9 · Output USD 3.6 · Cache read USD 0.18

The official model page lists 256K context; Standard API input/cached/output are $0.50/$0.05/$1.50 per MTok, with separate Batch and Priority tiers.

• StandardInput USD 0.5 · Output USD 1.5 · Cache read USD 0.05

The official model page lists 256K context; Standard API input/cached/output are $1.50/$0.15/$7.50 per MTok, with separate Batch and Priority tiers.

• StandardInput USD 1.5 · Output USD 7.5 · Cache read USD 0.15

The official model page lists 256K context; Standard API input/cached/output are $0.15/$0.015/$0.60 per MTok, with separate Batch and Priority tiers.

• StandardInput USD 0.15 · Output USD 0.6 · Cache read USD 0.015

The official model list and current chat pricing page were checked; this exact model is Deprecated, while current pricing lists K2.6, K2.7 Code, and K3. No input/output/cache rate was verified, and no nearby model was substituted.

Kimi K2.6 global API pricing is $0.95 input, $0.16 cached input, and $4 output per MTok; no cache-write price is listed.

• StandardInput USD 0.95 · Output USD 4 · Cache read USD 0.16

Kimi K2.7 Code global API pricing is $0.95 input, $0.19 cached input, and $4 output per MTok; no cache-write price is listed.

• StandardInput USD 0.95 · Output USD 4 · Cache read USD 0.19

Kimi K3 input/cached/output are $3/$0.30/$15 per MTok; default cache writes are $3/M for 5 minutes or $6/M for 1 hour, and a hit refreshes the TTL.

• StandardInput USD 3 · Output USD 15 · Cache read USD 0.3

The official pinned ID is grok-4.20-0309-multi-agent; at ≥200K input the whole request uses long-context rates, and Batch is 80% of Standard.

• Short context (<200K)Input USD 1.25 · Output USD 2.5 · Cache read USD 0.2
• Long context (≥200K, entire request)Input USD 2.5 · Output USD 5 · Cache read USD 0.4

The official pinned ID is grok-4.20-0309-non-reasoning; at ≥200K input the whole request uses long-context rates, and Batch is 80% of Standard.

• Short context (<200K)Input USD 1.25 · Output USD 2.5 · Cache read USD 0.2
• Long context (≥200K, entire request)Input USD 2.5 · Output USD 5 · Cache read USD 0.4

The official pinned ID is grok-4.20-0309-reasoning; at ≥200K input the whole request uses long-context rates, and Batch is 80% of Standard.

• Short context (<200K)Input USD 1.25 · Output USD 2.5 · Cache read USD 0.2
• Long context (≥200K, entire request)Input USD 2.5 · Output USD 5 · Cache read USD 0.4

At ≥200K input, the entire request uses long-context rates; Batch is 80% of Standard (20% discount).

• Short context (<200K)Input USD 1.25 · Output USD 2.5 · Cache read USD 0.2
• Long context (≥200K, entire request)Input USD 2.5 · Output USD 5 · Cache read USD 0.4

At ≥200K input, the entire request uses long-context rates; xAI does not publish a separate maximum output limit.

• Short context (<200K)Input USD 2 · Output USD 6 · Cache read USD 0.3
• Long context (≥200K, entire request)Input USD 4 · Output USD 12 · Cache read USD 0.6

At ≥200K input, the entire request uses long-context rates; the US regional endpoint adds 10% to all rates.

• Short context (<200K)Input USD 2 · Output USD 6 · Cache read USD 0.5
• Long context (≥200K, entire request)Input USD 4 · Output USD 12 · Cache read USD 1

At ≥200K input, the entire request uses long-context rates; the US regional endpoint adds 10% to all rates. Fast 2x is only for Cursor/Grok Build, not the public API.

• Short context (<200K)Input USD 2 · Output USD 6 · Cache read USD 0.5
• Long context (≥200K, entire request)Input USD 4 · Output USD 12 · Cache read USD 1

At ≥200K input, the entire request uses long-context rates; xAI does not publish a separate maximum output limit.

• Short context (<200K)Input USD 1 · Output USD 2 · Cache read USD 0.2
• Long context (≥200K, entire request)Input USD 2 · Output USD 4 · Cache read USD 0.4

Overseas rates are USD; mainland China real-time rates are ¥1/¥2 with ¥0.02/M cached input. Batch is half real-time; cache writes are free for a limited time. The pricing page does not state a context window.

• Real-time (overseas USD)Input USD 0.14 · Output USD 0.28 · Cache read USD 0.0028
• Batch (overseas USD)Input USD 0.07 · Output USD 0.14 · Cache read USD 0.0014

Overseas rates are USD; mainland China real-time rates are ¥3/¥6 with ¥0.025/M cached input. Batch is half real-time; cache writes are free for a limited time (recorded as 0). The pricing page does not state a context window.

• Real-time (overseas USD)Input USD 0.435 · Output USD 0.87 · Cache read USD 0.0036
• Batch (overseas USD)Input USD 0.2175 · Output USD 0.435 · Cache read USD 0.0018

Overseas rates are USD; mainland China real-time rates are ¥30/¥60 with ¥0.25/M cached input; Batch is unsupported. Cache writes are free for a limited time, and the pricing page does not state a context window.

• StandardInput USD 4.35 · Output USD 8.7 · Cache read USD 0.036

The exact model page and current official pricing page were checked. The model page lists 200K context, 128K maximum output, and model id glm-5-turbo; the pricing table lists only GLM-5.1, GLM-5, GLM-4.7, and GLM-4.6, so input/output/cache rates are unconfirmed.

GLM-5.1 input/cached/output are $1.40/$0.26/$4.40 per MTok; context is 200K, cache storage is free for a limited time, and no cache-write price is listed.

• StandardInput USD 1.4 · Output USD 4.4 · Cache read USD 0.26

GLM-5.2 input/cached/output are $1.40/$0.26/$4.40 per MTok; cache storage is free for a limited time and no cache-write unit price is listed.

• StandardInput USD 1.4 · Output USD 4.4 · Cache read USD 0.26

GLM-5.3 input/cached/output are $1.40/$0.26/$4.40 per MTok; cache storage is free for a limited time and no cache-write unit price is listed.

• StandardInput USD 1.4 · Output USD 4.4 · Cache read USD 0.26

GLM-5.3-Flash input/cached/output are $0.15/$0.03/$0.50 per MTok; cache storage is free for a limited time and no cache-write unit price is listed.

• StandardInput USD 0.15 · Output USD 0.5 · Cache read USD 0.03

GLM-5.3-FlashX input/cached/output are $0.37/$0.075/$1.25 per MTok; cache storage is free for a limited time and no cache-write unit price is listed.

• StandardInput USD 0.37 · Output USD 1.25 · Cache read USD 0.075

The exact multimodal model page and current official pricing page were checked. The model page lists 200K context and 128K maximum output; the pricing table has no GLM-5V-Turbo row, so input/output/cache rates are unconfirmed.