I checked the billing classification of Bedrock's Grok 4.6 in Cost Explorer
This page has been translated by machine translation. View original
Introduction
On August 19, 2026, Amazon Bedrock added support for SpaceXAI's Grok 4.6.
Amazon Bedrock now supports SpaceXAI Grok 4.6, a frontier model built for coding, agentic tasks, and knowledge work, with US Geo and Global cross-Region inference.
The US Geo inference profile—us.xai.grok-4.6—routes requests only within the US geography, so you can scale while keeping data processed within the United States to meet data residency requirements.
The inference profile IDs are us.xai.grok-4.6 for US Geo and global.xai.grok-4.6 for Global.
SpaceXAI has announced that Grok 4.6 matches high-end models across multiple benchmarks for agentic coding and knowledge work.
Grok 4.6 achieves frontier intelligence across several agentic coding and knowledge work benchmarks. It matches GPT-5.6 Sol on the Artificial Analysis Intelligence Index, which is a composite score of nine benchmarks.
IAM authentication on Bedrock, invocation via bedrock-mantle, and pricing comparisons are covered in the following article.
In this article, I verified which service and billing entity Grok 4.6 usage costs are categorized under in Cost Explorer. I also used the same dimensions to check how Claude and OpenAI GPT costs are categorized when called via Bedrock, and whether AWS credits are applied as offsets.
What Was Verified
The dimensions used were SERVICE, USAGE_TYPE, BILLING_ENTITY, and RECORD_TYPE.
Service Classification
Using a working account with no credits, I retrieved the services (SERVICE dimension values) that incurred costs between January 1, 2026 and September 1, 2026, and extracted only those related to Bedrock.
aws ce get-dimension-values \
--time-period Start=2026-01-01,End=2026-09-02 \
--dimension SERVICE \
| jq -r '.DimensionValues[].Value' | grep -E 'Bedrock'
Amazon Bedrock
Amazon Bedrock AgentCore
Claude 3 Haiku (Amazon Bedrock Edition)
Claude 3.5 Haiku (Amazon Bedrock Edition)
Claude 3.5 Sonnet v2 (Amazon Bedrock Edition)
Claude Fable 5 (Amazon Bedrock Edition)
Claude Haiku 4.5 (Amazon Bedrock Edition)
Claude Opus 4.7 (Amazon Bedrock Edition)
Claude Opus 4.8 (Amazon Bedrock Edition)
Claude Opus 5 (Amazon Bedrock Edition)
Claude Sonnet 4.5 (Amazon Bedrock Edition)
Claude Sonnet 4.6 (Amazon Bedrock Edition)
Claude Sonnet 5 (Amazon Bedrock Edition)
OpenAI GPT-5.4 (Amazon Bedrock Edition)
OpenAI GPT-5.5 (Amazon Bedrock Edition)
OpenAI GPT-5.6 Luna (Amazon Bedrock Edition)
OpenAI GPT-5.6 Sol (Amazon Bedrock Edition)
OpenAI GPT-5.6 Terra (Amazon Bedrock Edition)
Claude and OpenAI GPT appeared as independent services with "(Amazon Bedrock Edition)" appended per model. On the other hand, there is no Grok-specific entry among the SERVICE values containing "Bedrock" in their name. Grok costs are included under Amazon Bedrock. The next section verifies the USAGE_TYPE for this.
USAGE_TYPE
I filtered by SERVICE to Amazon Bedrock, aggregated by USAGE_TYPE, and extracted only the values for Grok, gpt-oss, and Gemma 4.
aws ce get-cost-and-usage \
--time-period Start=2026-01-01,End=2026-09-02 \
--granularity MONTHLY --metrics UnblendedCost \
--filter '{"Dimensions":{"Key":"SERVICE","Values":["Amazon Bedrock"]}}' \
--group-by '[{"Type":"DIMENSION","Key":"USAGE_TYPE"}]' \
| jq -r '[.ResultsByTime[].Groups[].Keys[0]]|unique|.[]' \
| grep -E 'grok|gpt-oss-120b|gemma-4-31b'
APN1-xai.grok-4.6-mantle-input-tokens-global-standard
APN1-xai.grok-4.6-mantle-output-tokens-global-standard
USE1-google.gemma-4-31b-mantle-input-tokens-standard
USE1-google.gemma-4-31b-mantle-output-tokens-standard
USE1-openai.gpt-oss-120b-mantle-input-tokens-standard
USE1-openai.gpt-oss-120b-mantle-output-tokens-standard
USE1-xai.grok-4.3-mantle-input-tokens-standard
USE1-xai.grok-4.3-mantle-output-tokens-standard
USE1-xai.grok-4.6-mantle-input-tokens-standard
USE1-xai.grok-4.6-mantle-output-tokens-standard
USE2-gpt-oss-120b-input-tokens
USE2-gpt-oss-120b-output-tokens
USE2-openai.gpt-oss-120b-mantle-input-tokens-standard
USE2-openai.gpt-oss-120b-mantle-output-tokens-standard
USW2-google.gemma-4-31b-mantle-input-tokens-standard
USW2-google.gemma-4-31b-mantle-output-tokens-standard
USW2-openai.gpt-oss-120b-mantle-input-tokens-standard
USW2-openai.gpt-oss-120b-mantle-output-tokens-standard
USW2-xai.grok-4.3-mantle-cache-read-tokens-standard
USW2-xai.grok-4.3-mantle-input-tokens-standard
USW2-xai.grok-4.3-mantle-output-tokens-standard
USW2-xai.grok-4.6-mantle-cache-read-tokens-standard
USW2-xai.grok-4.6-mantle-input-tokens-standard
USW2-xai.grok-4.6-mantle-output-tokens-standard
The USAGE_TYPE for Grok 4.6 follows the pattern <region-prefix>-xai.grok-4.6-mantle-<input|output|cache-read>-tokens-standard. This follows the same naming convention as the mantle-type values for gpt-oss 120B and Gemma 4 31B, and all are categorized under Amazon Bedrock. An alternative format without the provider name and mantle also coexists, as seen with USE2-gpt-oss-120b-input-tokens in the list. For cross-region inference via the global profile, the USAGE_TYPE suffix becomes global-standard. In the list, APN1-xai.grok-4.6-mantle-input-tokens-global-standard corresponds to this.
The full output of the above command without the grep filter also included USAGE_TYPE values for Qwen, DeepSeek, Mistral, MiniMax, Kimi, Nemotron, GLM, and Nova. Claude and OpenAI GPT-5.x USAGE_TYPE values did not appear.
Billing Entity
In this aggregation, even models invoked via Bedrock were split between AWS and AWS Marketplace as billing entities.
Grok 4.6 Billing Entity Is AWS
For August 2026, I aggregated by BILLING_ENTITY and RECORD_TYPE for the 7 USAGE_TYPE values corresponding to Grok 4.6. As in the previous section, this is from a working account with no credits. The billing entity was AWS, and the only rows in the results had a RECORD_TYPE of Usage. Since the billing amount was not yet finalized at the time of retrieval, the values are estimates.
Grok 4.6 billing entity (full command and output)
aws ce get-cost-and-usage \
--time-period Start=2026-08-01,End=2026-09-01 \
--granularity MONTHLY --metrics UnblendedCost \
--filter '{"Dimensions":{"Key":"USAGE_TYPE","Values":["APN1-xai.grok-4.6-mantle-input-tokens-global-standard","APN1-xai.grok-4.6-mantle-output-tokens-global-standard","USE1-xai.grok-4.6-mantle-input-tokens-standard","USE1-xai.grok-4.6-mantle-output-tokens-standard","USW2-xai.grok-4.6-mantle-cache-read-tokens-standard","USW2-xai.grok-4.6-mantle-input-tokens-standard","USW2-xai.grok-4.6-mantle-output-tokens-standard"]}}' \
--group-by '[{"Type":"DIMENSION","Key":"BILLING_ENTITY"},{"Type":"DIMENSION","Key":"RECORD_TYPE"}]'
{
"GroupDefinitions": [
{
"Type": "DIMENSION",
"Key": "BILLING_ENTITY"
},
{
"Type": "DIMENSION",
"Key": "RECORD_TYPE"
}
],
"ResultsByTime": [
{
"TimePeriod": {
"Start": "2026-08-01",
"End": "2026-09-01"
},
"Total": {},
"Groups": [
{
"Keys": [
"AWS",
"Usage"
],
"Metrics": {
"UnblendedCost": {
"Amount": "0.0182128794",
"Unit": "USD"
}
}
}
],
"Estimated": true
}
],
"DimensionValueAttributes": []
}
Claude Opus 5 and OpenAI GPT-5.6 Billing Entity Is AWS Marketplace
For August 2026 on this account, I filtered by AWS Marketplace as the billing entity and aggregated by service.
aws ce get-cost-and-usage \
--time-period Start=2026-08-01,End=2026-09-01 \
--granularity MONTHLY --metrics UnblendedCost \
--filter '{"Dimensions":{"Key":"BILLING_ENTITY","Values":["AWS Marketplace"]}}' \
--group-by '[{"Type":"DIMENSION","Key":"SERVICE"},{"Type":"DIMENSION","Key":"RECORD_TYPE"}]' \
| jq -r '.ResultsByTime[].Groups[]|[.Keys[0],.Keys[1],.Metrics.UnblendedCost.Amount]|@tsv'
Claude Opus 5 (Amazon Bedrock Edition) Usage 0.0000715
OpenAI GPT-5.6 Luna (Amazon Bedrock Edition) Usage 1.4188744473
OpenAI GPT-5.6 Sol (Amazon Bedrock Edition) Usage 0.002095
OpenAI GPT-5.6 Terra (Amazon Bedrock Edition) Usage 0.08698026
Tax Tax 0
The AWS Marketplace aggregation listed Claude Opus 5 and OpenAI GPT-5.6 (Luna / Sol / Terra), each invoked via Bedrock, as separate services. Grok 4.6 was not included in this list.
Credit Offset
In addition to the working account with no credits, I also aggregated August 2026 data for an account that holds AWS credits and used Grok 4.6 in us-west-2. The scope was SERVICE = Amazon Bedrock, with the 2 USAGE_TYPE values corresponding to Grok 4.6 input and output tokens. The aggregation was done by RECORD_TYPE and USAGE_TYPE. Since the billing amount was not yet finalized at the time of retrieval, the values are estimates.
Credit USW2-xai.grok-4.6-mantle-input-tokens-standard -0.0000000836
Credit USW2-xai.grok-4.6-mantle-output-tokens-standard -0.000002046
Usage USW2-xai.grok-4.6-mantle-input-tokens-standard 0.0000000836
Usage USW2-xai.grok-4.6-mantle-output-tokens-standard 0.000002046
For both input and output tokens, a Credit row equal to the Usage amount was recorded, resulting in a total of 0. When the same data was aggregated by billing entity, the AWS Usage and Credit amounts were also equal. Credits were applied to Grok 4.6 usage costs billed under AWS.
The credits I verified in this case included Amazon Bedrock as an eligible service. You can check which services your AWS credits apply to in the credits list in the Billing console.
Summary
If Amazon Bedrock is included in the eligible services for your AWS credits, credits can also be applied to Grok 4.6 usage costs. If you are looking for a high-performance LLM you can use with credits, give Grok 4.6 a try as one of your options.
