fix(pricing): apply AWS GPT-5.6 Sol Bedrock rate cut

[lmcdonald-godaddy]

Summary

AWS cut GPT-5.6 Sol on Bedrock on Aug 21 (Global CRIS $4 / $20 per 1M tokens; Geo/In-Region $4.40 / $22). #37880 already applied the matching OpenAI promo and explicitly left Bedrock at the old card. The live Sol model card now has the new rates, so Mantle and Converse still overstate spend.

This PR only rebases:

• bedrock_mantle/openai.gpt-5.6-sol and us.openai.gpt-5.6-sol to Geo/In-Region ($4.40 / $22, cache write $5.50, cache read $0.44)
• global.openai.gpt-5.6-sol to Global CRIS ($4 / $20, cache write $5.00, cache read $0.40)

Long-context (>272k) fields follow the card (2x input, 1.5x output). Terra, Luna, Azure, and OpenAI entries are untouched.

Relevant issues

Follow-up to #37880. Source: https://aws.amazon.com/about-aws/whats-new/2026/08/bedrock-openai-gpt-56-sol-reduced-pricing/ and https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-openai-gpt-56-sol.html

Linear ticket

Pre-Submission checklist

• I have added meaningful tests
• My PR passes all unit tests on make test-unit
• My PR's scope is as isolated as possible; it only solves 1 specific problem
• I have requested a Greptile review by commenting @greptileai and received a Confidence Score of at least 4/5 before requesting a maintainer review

Screenshots / Proof of Fix

Pinning tests now expect the Aug 21 rates. After merge, with a local proxy and LITELLM_LOCAL_MODEL_COST_MAP=True:

curl -sS -D - -X POST http://localhost:4000/v1/chat/completions \
  -H "Authorization: Bearer sk-1234" -H "Content-Type: application/json" \
  -d '{"model":"bedrock/converse/us.openai.gpt-5.6-sol","messages":[{"role":"user","content":"Reply with the single word: ok"}],"max_tokens":16}'

x-litellm-response-cost should match prompt_tokens * 4.4e-6 + completion_tokens * 2.2e-5 (Geo). For bedrock/converse/global.openai.gpt-5.6-sol use 4e-6 and 2e-5. Same check on bedrock_mantle/openai.gpt-5.6-sol via /v1/responses.

Type

Bug Fix

Changes

Cost map Sol Bedrock entries and the tests that pin them

Made with Cursor

[greptile-apps]

Greptile Summary

Updates GPT-5.6 Sol pricing metadata for AWS Bedrock’s reduced Global and Geo/In-Region rates, including cache and long-context tiers. Corresponding tests now pin the revised rates.

• Applies Geo/In-Region pricing to Bedrock Mantle and the US inference profile.
• Applies Global CRIS pricing to the global inference profile.
• Keeps the primary and packaged backup cost maps synchronized.
• Updates Bedrock Converse and Mantle pricing tests without weakening their behavioral assertions.

Confidence Score: 5/5

The PR appears safe to merge; the revised pricing entries are consistently mapped, mirrored, and covered by independent assertions.

The changed Global, US, and Mantle model identifiers resolve to their intended pricing entries, and the updated tests exercise the affected standard, cache, and long-context calculations without revealing a blocking failure.

Important Files Changed

Filename	Overview
model_prices_and_context_window.json	Updates the three intended GPT-5.6 Sol Bedrock entries with consistent base, cache, and long-context rates.
litellm/model_prices_and_context_window_backup.json	Mirrors the primary pricing-map changes in the packaged backup used by local model-cost resolution.
tests/test_litellm/llms/bedrock/test_cross_region_inference_profile_mapping.py	Independently pins revised US and global rates and continues exercising normal, cache, and long-context billing.
tests/test_litellm/llms/bedrock_mantle/test_bedrock_mantle_responses_transformation.py	Updates independent Mantle pricing expectations while retaining coverage of standard and long-context cost behavior.

_{Reviews (1): Last reviewed commit: "fix(pricing): apply AWS GPT-5.6 Sol Bedr..." | Re-trigger Greptile}

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

[codspeed-hq]

Merging this PR will not alter performance

✅ 31 untouched benchmarks

---

_{Comparing lmcdonald-godaddy:litellm_bedrock_gpt56_sol_promo (44a93e9) with litellm_internal_staging (d447be1)¹}

Footnotes

1. No successful run was found on litellm_internal_staging (a626170) during the generation of this report, so d447be1 was used instead as the comparison base. There might be some changes unrelated to this pull request in this report. ↩