Skip to content

feat: add OpenAI GPT-5.6 Sol/Terra/Luna pricing with context tiers for Bedrock CRIS - #891

Merged
narengogi merged 1 commit into
mainfrom
support/8200-bedrock-cris-openai
Aug 21, 2026
Merged

feat: add OpenAI GPT-5.6 Sol/Terra/Luna pricing with context tiers for Bedrock CRIS#891
narengogi merged 1 commit into
mainfrom
support/8200-bedrock-cris-openai

Conversation

@akhilmmenon

@akhilmmenon akhilmmenon commented Aug 19, 2026

Copy link
Copy Markdown
Member

Summary

  • Add pricing and general config entries for OpenAI GPT-5.6 Sol, Terra, and Luna models on AWS Bedrock (bedrock-runtime endpoint) with cross-region inference profiles
  • 8 model IDs added: us. (Geo CRIS), in. (Geo CRIS, Terra/Luna only), global. (Global CRIS)
  • Pricing includes custom_pricing with context_tier_map for Short (≤272K) and Long (>272K) context window tiers, following the same structure as vertex-ai.json

Models added to pricing/bedrock.json:

  • us.openai.gpt-5.6-sol, global.openai.gpt-5.6-sol
  • us.openai.gpt-5.6-terra, in.openai.gpt-5.6-terra, global.openai.gpt-5.6-terra
  • us.openai.gpt-5.6-luna, in.openai.gpt-5.6-luna, global.openai.gpt-5.6-luna

Models added to general/bedrock.json:

  • Same 8 model IDs with Converse API-aligned config (max_tokens: 272000, image support, temperature/top_p removed as reasoning models)

Pricing structure

  • pricing_config (top-level): Short context fallback pricing
  • custom_pricing.context_tier_map: {"0": "lte-272k", "272000": "gt-272k"}
  • custom_pricing.regions.default.execution_modes.standard.context_tiers: Short and Long tier pricing

Region-based pricing (custom_pricing.regions) uses "default" because these models are only available via cross-region inference profiles on bedrock-runtime — AWS charges a flat rate per inference option (Geo/Global), not per individual AWS region. The Geo vs Global price difference is encoded in separate model IDs (us./in. vs global.).

Pricing source

Ref: Pylon #8200

Test plan

  • Pricing values verified against AWS model cards for all 3 models × all inference options × both context tiers
  • Structure consistent with existing vertex-ai.json context tier pattern
  • [] E2E tested with live Bedrock requests (200 OK responses with correct model routing)

Copilot AI left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Pull request overview

Adds Bedrock cross-region pricing and model metadata for eight OpenAI GPT-5.6 inference profiles.

Changes:

  • Adds geo and global pricing for Sol, Terra, and Luna.
  • Adds token limits, image support, and reasoning-model parameter exclusions.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 3 comments.

File Description
pricing/bedrock.json Adds cross-region GPT-5.6 pricing.
general/bedrock.json Adds GPT-5.6 model configurations.
Suppressed comments (5)

general/bedrock.json:2122

  • This profile omits the model's tool-use capability, so consumers of this metadata will disable tools even though the same model is marked tool-capable in general/bedrock-mantle.json:353-356. Include tools in the supported capabilities for Converse requests.
    "type": { "primary": "chat", "supported": ["image"] },

general/bedrock.json:2132

  • This profile omits the model's tool-use capability, so consumers of this metadata will disable tools even though the same model is marked tool-capable in general/bedrock-mantle.json:358-361. Include tools in the supported capabilities for Converse requests.
    "type": { "primary": "chat", "supported": ["image"] },

general/bedrock.json:2137

  • This profile omits the model's tool-use capability, so consumers of this metadata will disable tools even though the same model is marked tool-capable in general/bedrock-mantle.json:358-361. Include tools in the supported capabilities for Converse requests.
    "type": { "primary": "chat", "supported": ["image"] },

general/bedrock.json:2147

  • This profile omits the model's tool-use capability, so consumers of this metadata will disable tools even though the same model is marked tool-capable in general/bedrock-mantle.json:363-366. Include tools in the supported capabilities for Converse requests.
    "type": { "primary": "chat", "supported": ["image"] },

general/bedrock.json:2152

  • This profile omits the model's tool-use capability, so consumers of this metadata will disable tools even though the same model is marked tool-capable in general/bedrock-mantle.json:363-366. Include tools in the supported capabilities for Converse requests.
    "type": { "primary": "chat", "supported": ["image"] },

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

Comment thread general/bedrock.json
Comment thread general/bedrock.json
Comment thread general/bedrock.json
@akhilmmenon

Copy link
Copy Markdown
Member Author

Test Results

Config validation

All 8 model entries resolve correctly in both pricing and general configs:

--- Pricing ---
✓ us.openai.gpt-5.6-sol {"price":0.00055}
✓ global.openai.gpt-5.6-sol {"price":0.0005}
✓ us.openai.gpt-5.6-terra {"price":0.00022}
✓ in.openai.gpt-5.6-terra {"price":0.00022}
✓ global.openai.gpt-5.6-terra {"price":0.0002}
✓ us.openai.gpt-5.6-luna {"price":0.000022}
✓ in.openai.gpt-5.6-luna {"price":0.000022}
✓ global.openai.gpt-5.6-luna {"price":0.00002}

--- General ---
✓ us.openai.gpt-5.6-sol [ 'image' ]
✓ global.openai.gpt-5.6-sol [ 'image' ]
✓ us.openai.gpt-5.6-terra [ 'image' ]
✓ in.openai.gpt-5.6-terra [ 'image' ]
✓ global.openai.gpt-5.6-terra [ 'image' ]
✓ us.openai.gpt-5.6-luna [ 'image' ]
✓ in.openai.gpt-5.6-luna [ 'image' ]
✓ global.openai.gpt-5.6-luna [ 'image' ]

End-to-end verification

Tested locally against bedrock-runtime with the gateway regex fix applied. Request to us.openai.gpt-5.6-luna returns 200 OK with correct model attribution and usage tracking:

{
    "model": "us.openai.gpt-5.6-luna",
    "provider": "bedrock",
    "choices": [{"message": {"content": "Hello! How can I help you today?"}, "finish_reason": "stop"}],
    "usage": {"prompt_tokens": 8, "completion_tokens": 13, "total_tokens": 21}
}

narengogi
narengogi previously approved these changes Aug 20, 2026
@narengogi

Copy link
Copy Markdown
Member

@akhilmmenon please resolve conflicts

…r openai-gpt-5.6-sol/terra/luna models in bedrock-runtime

Co-authored-by: Cursor <cursoragent@cursor.com>
@akhilmmenon
akhilmmenon force-pushed the support/8200-bedrock-cris-openai branch from f5697b9 to 2169346 Compare August 20, 2026 08:51
@akhilmmenon akhilmmenon changed the title feat(bedrock): add OpenAI GPT-5.6 cross-region inference pricing and configs feat: add OpenAI GPT-5.6 Sol/Terra/Luna pricing with context tiers for Bedrock CRIS Aug 20, 2026
@narengogi
narengogi requested a balanced review from Copilot August 21, 2026 03:05

Copilot AI left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Pull request overview

Copilot reviewed 2 out of 2 changed files in this pull request and generated no new comments.

Suppressed comments (8)

general/bedrock.json:2121

  • max_tokens caps generated output, but GPT-5.6 Sol's output limit is 128,000 (as also recorded in general/openai.json:2902 and general/azure-openai.json:4252). Using the 272,000 input-pricing threshold here exposes values that Bedrock cannot accept; cap this at 128,000.
    "params": [{ "key": "max_tokens", "maxValue": 272000 }],

general/bedrock.json:2131

  • max_tokens caps generated output, but GPT-5.6 Terra's output limit is 128,000 (as also recorded in general/openai.json:2944 and general/azure-openai.json:4312). Using the 272,000 input-pricing threshold here exposes values that Bedrock cannot accept; cap this at 128,000.
    "params": [{ "key": "max_tokens", "maxValue": 272000 }],

general/bedrock.json:2136

  • max_tokens caps generated output, but GPT-5.6 Terra's output limit is 128,000 (as also recorded in general/openai.json:2944 and general/azure-openai.json:4312). Using the 272,000 input-pricing threshold here exposes values that Bedrock cannot accept; cap this at 128,000.
    "params": [{ "key": "max_tokens", "maxValue": 272000 }],

general/bedrock.json:2151

  • max_tokens caps generated output, but GPT-5.6 Luna's output limit is 128,000 (as also recorded in general/openai.json:2986 and general/azure-openai.json:4372). Using the 272,000 input-pricing threshold here exposes values that Bedrock cannot accept; cap this at 128,000.
    "params": [{ "key": "max_tokens", "maxValue": 272000 }],

general/bedrock.json:2146

  • max_tokens caps generated output, but GPT-5.6 Luna's output limit is 128,000 (as also recorded in general/openai.json:2986 and general/azure-openai.json:4372). Using the 272,000 input-pricing threshold here exposes values that Bedrock cannot accept; cap this at 128,000.
    "params": [{ "key": "max_tokens", "maxValue": 272000 }],

general/bedrock.json:2116

  • max_tokens caps generated output, but GPT-5.6 Sol's output limit is 128,000 (as also recorded in general/openai.json:2902 and general/azure-openai.json:4252). Using the 272,000 input-pricing threshold here exposes values that Bedrock cannot accept; cap this at 128,000.

This issue also appears on line 2121 of the same file.

    "params": [{ "key": "max_tokens", "maxValue": 272000 }],

general/bedrock.json:2126

  • max_tokens caps generated output, but GPT-5.6 Terra's output limit is 128,000 (as also recorded in general/openai.json:2944 and general/azure-openai.json:4312). Using the 272,000 input-pricing threshold here exposes values that Bedrock cannot accept; cap this at 128,000.

This issue also appears in the following locations of the same file:

  • line 2131
  • line 2136
    "params": [{ "key": "max_tokens", "maxValue": 272000 }],

general/bedrock.json:2141

  • max_tokens caps generated output, but GPT-5.6 Luna's output limit is 128,000 (as also recorded in general/openai.json:2986 and general/azure-openai.json:4372). Using the 272,000 input-pricing threshold here exposes values that Bedrock cannot accept; cap this at 128,000.

This issue also appears in the following locations of the same file:

  • line 2146
  • line 2151
    "params": [{ "key": "max_tokens", "maxValue": 272000 }],

@narengogi
narengogi merged commit 90d5ab3 into main Aug 21, 2026
3 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants