chore(prices): sync AWS Bedrock prices: 6 models [enrichment failed: AWS Bedrock, 4 held]

[berriai-litellm-provider-info-sync]

TLDR

Problem this solves:

• Bedrock catalog entries drifted from the published pricing page
• Six models carried stale token limits and capability flags

How it solves it:

• Syncs 6 AWS Bedrock models from the provider page
• Only model_prices_and_context_window.json and its copies change
• Four unverified kimi-k3 values held out of the diff

User Flow

A before and after user flow does not apply to an automated price sync; the only change is updated metadata in the model catalog file

Relevant issues

Affected release

Linear ticket

Pre-Submission checklist

Please complete all items before asking a LiteLLM maintainer to review your PR

• I have added meaningful tests
• The handful of test files covering my change pass locally, e.g. uv run pytest tests/test_litellm/<your_test_file>.py -v. Leave the suites (make test-unit-*, make test-unit) to CI: it finishes in ~15 minutes where a laptop takes an hour or more
• My PR passes all required CI/CD checks (e.g., lint, schema.d.ts sync check, etc.)
• My PR's scope is as isolated as possible; it only solves 1 specific problem
• I have received a Greptile Confidence Score of at least 4/5 before requesting a maintainer review (Greptile reviews automatically once the PR is opened; only comment @greptileai to re-request a review after pushing changes)

Delays in PR merge?

If you're seeing a delay in your PR being merged, ping the LiteLLM Team on Slack (#pr-review).

Screenshots / Proof of Fix

Screenshots and live command output do not apply to an automated price sync; each applied value was read from the AWS Bedrock pricing page and the six models were verified through LiteLLM on /v1/chat/completions before being written

Type

Chore

Caveats (if any)

Medium

• Four kimi-k3 prices held, need manual review and apply
  • LiteLLM probe returned 400: on-demand throughput unsupported, inference profile required
  • Page lists $3/1M in, $15/1M out, $0.30/1M cache read, $3.75/1M cache write
• minimax.minimax-m2.5 max_input_tokens drops from 1T to 196B

Low

• This run changes token limits and capability flags, not per-token prices
• Enrichment failed for AWS Bedrock on this run
• 28 priced rows skipped: the page lists no Bedrock model id

QA runbook

This PR does not add or change tests, so there is no manual runbook

Final Attestation

• The tests check the right things, including the edge cases, and regressions in the respective real-world customer use-cases are not possible after this PR

Sync report

Updates 6 models, each read from its provider's own published prices.

AWS Bedrock

Updates 6 models from the provider's pricing page (page c21fff8ee033).

⚠️ New models held for review. The run could not verify one or more new models (enrichment, the free-model check, or a LiteLLM probe), so its catalog entry remains unchanged until the check passes. The reason follows.

new models held pending a LiteLLM probe: moonshotai.kimi-k3 (LiteLLM returned HTTP 400 (invalid_request_error): "litellm.BadRequestError: BedrockException - {\"message\":\"Invocation of model ID moonshotai.kimi-k3 with on-demand throughput isn’t supported. Retry your request with the ID or ARN of an inference profile that contains this model.\"}No fallback model group found for lookup_groups=bedrock/moonshotai.kim…")

Verified through LiteLLM before being added: 6 via /v1/chat/completions.

Changes

• minimax.minimax-m2: max_input_tokens $128000000000/1M → $1000000000000/1M, max_output_tokens $8192000000/1M → $8000000000/1M, max_tokens $8192000000/1M → $8000000000/1M, supports_audio_input false, supports_function_calling true, supports_response_schema true, supports_vision false, source set
• minimax.minimax-m2.1: max_output_tokens $8192000000/1M → $8000000000/1M, max_tokens $8192000000/1M → $8000000000/1M, supports_audio_input false, supports_response_schema true, supports_vision false
• minimax.minimax-m2.5: max_input_tokens $1000000000000/1M → $196000000000/1M, max_output_tokens $8192000000/1M → $8000000000/1M, max_tokens $8192000000/1M → $8000000000/1M, supports_audio_input false, supports_response_schema true, supports_vision false
• moonshot.kimi-k2-thinking: max_input_tokens $128000000000/1M → $256000000000/1M, max_output_tokens $8192000000/1M → $16000000000/1M, max_tokens $8192000000/1M → $16000000000/1M, supports_audio_input false, supports_function_calling true, supports_response_schema true, supports_vision false, source set
• openai.gpt-oss-safeguard-120b: max_output_tokens $8192000000/1M → $16000000000/1M, max_tokens $8192000000/1M → $16000000000/1M, supports_audio_input false, supports_function_calling true, supports_response_schema true, supports_vision false, source set
• openai.gpt-oss-safeguard-20b: max_output_tokens $8192000000/1M → $16000000000/1M, max_tokens $8192000000/1M → $16000000000/1M, supports_audio_input false, supports_function_calling true, supports_response_schema true, supports_vision false, source set

Held for review (4) — not in this diff

These values were read from the source but did not pass the gates on their own. A human needs to check them against the page and apply them by hand if right.

• moonshotai.kimi-k3 input_cost_per_token: $3/1M — smoke-failed: LiteLLM returned HTTP 400 (invalid_request_error): "litellm.BadRequestError: BedrockException - {"message":"Invocation of model ID moonshotai.kimi-k3 with on-demand throughput isn’t supported. Retry your request with the ID or ARN of an inference profile that contains this model."}No fallback model group found for lookup_groups=bedrock/moonshotai.kim…" (https://aws.amazon.com/bedrock/pricing/ row "Global CRIS" input_cost_per_token = $3.00 per 1M tokens)
• moonshotai.kimi-k3 output_cost_per_token: $15/1M — smoke-failed: LiteLLM returned HTTP 400 (invalid_request_error): "litellm.BadRequestError: BedrockException - {"message":"Invocation of model ID moonshotai.kimi-k3 with on-demand throughput isn’t supported. Retry your request with the ID or ARN of an inference profile that contains this model."}No fallback model group found for lookup_groups=bedrock/moonshotai.kim…" (https://aws.amazon.com/bedrock/pricing/ row "Global CRIS" output_cost_per_token = $15.00 per 1M tokens)
• moonshotai.kimi-k3 cache_read_input_token_cost: $0.3/1M — smoke-failed: LiteLLM returned HTTP 400 (invalid_request_error): "litellm.BadRequestError: BedrockException - {"message":"Invocation of model ID moonshotai.kimi-k3 with on-demand throughput isn’t supported. Retry your request with the ID or ARN of an inference profile that contains this model."}No fallback model group found for lookup_groups=bedrock/moonshotai.kim…" (https://aws.amazon.com/bedrock/pricing/ row "Global CRIS" cache_read_input_token_cost = $0.30 per 1M tokens)
• moonshotai.kimi-k3 cache_creation_input_token_cost: $3.75/1M — smoke-failed: LiteLLM returned HTTP 400 (invalid_request_error): "litellm.BadRequestError: BedrockException - {"message":"Invocation of model ID moonshotai.kimi-k3 with on-demand throughput isn’t supported. Retry your request with the ID or ARN of an inference profile that contains this model."}No fallback model group found for lookup_groups=bedrock/moonshotai.kim…" (https://aws.amazon.com/bedrock/pricing/ row "Global CRIS" cache_creation_input_token_cost = $3.75 per 1M tokens)

No litellm key or field (28)

Rows the page prices that litellm has no key or field for. Nothing to apply; listed so the omission is deliberate, not missed.

• Claude: not in the synced set: the price list states no Bedrock model id, and this model's id is not its display name
• Claude (100K): not in the synced set: the price list states no Bedrock model id, and this model's id is not its display name
• Claude 3 Haiku: not in the synced set: the price list states no Bedrock model id, and this model's id is not its display name
• Claude 3 Opus: not in the synced set: the price list states no Bedrock model id, and this model's id is not its display name
• Claude 3 Sonnet: not in the synced set: the price list states no Bedrock model id, and this model's id is not its display name
• Claude 3.5 Haiku: not in the synced set: the price list states no Bedrock model id, and this model's id is not its display name
• Claude 3.5 Sonnet: not in the synced set: the price list states no Bedrock model id, and this model's id is not its display name
• Claude 3.5 Sonnet v2: not in the synced set: the price list states no Bedrock model id, and this model's id is not its display name
• Claude 3.7 Sonnet: not in the synced set: the price list states no Bedrock model id, and this model's id is not its display name
• Claude Haiku 4.5: not in the synced set: the price list states no Bedrock model id, and this model's id is not its display name
• Claude Instant: not in the synced set: the price list states no Bedrock model id, and this model's id is not its display name
• Claude Instant (100K): not in the synced set: the price list states no Bedrock model id, and this model's id is not its display name
• Claude Mythos 5: not in the synced set: the price list states no Bedrock model id, and this model's id is not its display name
• Claude Mythos 5.1: not in the synced set: the price list states no Bedrock model id, and this model's id is not its display name
• Claude Mythos Preview: not in the synced set: the price list states no Bedrock model id, and this model's id is not its display name
• Claude Opus 4: not in the synced set: the price list states no Bedrock model id, and this model's id is not its display name
• Claude Opus 4.1: not in the synced set: the price list states no Bedrock model id, and this model's id is not its display name
• Claude Opus 4.5: not in the synced set: the price list states no Bedrock model id, and this model's id is not its display name
• Claude Opus 4.6: not in the synced set: the price list states no Bedrock model id, and this model's id is not its display name
• Claude Sonnet 4: not in the synced set: the price list states no Bedrock model id, and this model's id is not its display name
• Claude Sonnet 4.5: not in the synced set: the price list states no Bedrock model id, and this model's id is not its display name
• Gemma 4 26B A4B: not in the synced set: the page states no Bedrock model id
• Gemma 4 31B: not in the synced set: the page states no Bedrock model id
• Gemma 4 E2B: not in the synced set: the page states no Bedrock model id
• Llama 2 Chat (13B): not in the synced set: the page states no Bedrock model id
• Llama 2 Chat (70B): not in the synced set: the page states no Bedrock model id
• Voxtral Mini 1.0: not in the synced set: the page states no Bedrock model id
• Voxtral Small 1.0: not in the synced set: the page states no Bedrock model id

Opened by the litellm-providers price sync. Every price is read from the provider's own published source and gated before it is applied; the audit trail for each value is in the portal's sync history.

---

Note

Medium Risk
Catalog context and capability flags drive routing and validation; the large minimax.minimax-m2.5 input limit reduction could change which requests LiteLLM accepts for that model.

Overview
Syncs six AWS Bedrock bedrock_converse catalog entries in model_prices_and_context_window.json and its backup from the AWS Bedrock pricing page, without changing per-token prices in this diff.

Token limits are updated: minimax.minimax-m2 gains a much larger max_input_tokens (128k → 1M) with output capped at 8k; minimax.minimax-m2.5 corrects max_input_tokens down (1M → 196k) and aligns output to 8k; moonshot.kimi-k2-thinking doubles input context (128k → 256k) and raises max output (8k → 16k); both openai.gpt-oss-safeguard-* models get 16k max output instead of 8k.

Capability metadata is filled in across these models (supports_function_calling, supports_response_schema, supports_audio_input / supports_vision, and source URLs where missing). moonshotai.kimi-k3 pricing is intentionally not included (held for manual review per the sync report).

^{Reviewed by Cursor Bugbot for commit b305928. Bugbot is set up for automated code reviews on this repo. Configure here.}

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}

[greptile-apps]

Confidence Score: 5/5

The catalog-only update appears safe to merge with no concrete correctness, security, or synchronization defects identified

Summary

This PR synchronizes token limits, capability flags, and source metadata for six AWS Bedrock model entries. The root catalog and packaged backup remain byte-identical, and the updated token fields preserve the repository’s max-token consistency invariant

_{Reviews (1) · Last reviewed commit: "chore(prices): sync AWS Bedrock prices: ..."}

[codspeed]

Merging this PR will not alter performance

✅ 31 untouched benchmarks

---

_{Comparing litellm-providers/price-sync-aws-bedrock (b305928) with main (7b8bc54)}

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

[litellm-stampbot]

LGTM