[bhishmendramahala-crypto]

Problem

AWS Bedrock now supports an Ultrafast service tier for GPT-6 Astra, priced at 6× Standard. Without this, any request using service_tier: "ultrafast" falls back to Standard pricing — undercounting cost by 6×.

Source: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-openai-gpt-6-astra.html

Changes

pricing/bedrock.json

• us.openai.gpt-6-astra — added execution_modes.ultrafast (short + long context)
• global.openai.gpt-6-astra — added execution_modes.ultrafast (short + long context)
• openai.gpt-6-astra (bedrock-runtime In-Region) — added execution_modes.ultrafast (short + long context)

pricing/bedrock-mantle.json

• openai.gpt-6-astra — added full custom_pricing block with execution_modes.standard + execution_modes.ultrafast (previously had only flat pricing_config, so long-context standard was also missing)

Prices (USD per 1M tokens)

Model ID	Tier	Context	Input	Output	Cache Write	Cache Read
us.* / In-Region	ultrafast	≤272K	$66.00	$330.00	$82.50	$6.60
us.* / In-Region	ultrafast	>272K	$132.00	$495.00	$165.00	$13.20
global.*	ultrafast	≤272K	$60.00	$300.00	$75.00	$6.00
global.*	ultrafast	>272K	$120.00	$450.00	$150.00	$12.00

Pylon Ref : https://app.usepylon.com/issues?issueNumber=9044

Testing

• JSON validated via python3 -c "import json; json.load(open(...))"
• Prices cross-checked against AWS docs (6× standard confirmed)
• Companion gateway PR (type union): https://code.pan.run/airs/agent-gateway/backend/gateway-enterprise-node/-/merge_requests/2020

Notes

We are not using Ultrafast (SCP-denied) — this is purely for accurate cost tracking of any traffic that does reach this tier.

[Copilot]

🟡 Changes recommended

All four new context-tier configurations omit the model’s web-search and file-search fees, leaving tier-specific pricing incomplete.

1 open issue

• Add web and file search rates to context-tier configs · New

What changed in this PR

Adds GPT-6 Astra Ultrafast pricing to the Bedrock pricing catalog so cost tracking can distinguish it from Standard service.

Changes:

• Adds Ultrafast rates for US, global, and In-Region Bedrock model IDs across both context tiers.
• Adds Standard and Ultrafast context-tier pricing for Bedrock Mantle, including long-context Standard rates.

File	Description
pricing/​bedrock.json	Adds Ultrafast token and cache rates for three GPT-6 Astra model IDs.
pricing/​bedrock-mantle.json	Adds context-based Standard and Ultrafast pricing for GPT-6 Astra.

_{Review effort: Balanced}