fix: apply cache_read_input_token_cost to cached tokens in Fireworks AI cost calculation

[VANDRANKI]

What

litellm/llms/fireworks_ai/cost_calculator.py ignores cached_tokens when computing prompt cost. All input tokens are charged at input_cost_per_token even when cache_read_input_token_cost is configured and cache hits are reported in prompt_tokens_details.

Fixes #25950.

Why

The cost_per_token function did:

prompt_cost = usage["prompt_tokens"] * model_info["input_cost_per_token"]

This charges every token at the full input rate, ignoring cached_tokens in prompt_tokens_details and the cache_read_input_token_cost / cache_creation_input_token_cost fields that are correctly loaded into model_info.

Fix

Extract cached_tokens and cache_creation_tokens from usage.prompt_tokens_details, then apply the appropriate rate to each bucket:

non_cached_tokens * input_cost_per_token
+ cached_tokens * cache_read_input_token_cost
+ cache_creation_tokens * cache_creation_input_token_cost

Falls back to 0.0 if the cache cost fields are not set, so standard Fireworks serverless tier pricing is unchanged.

Test

Using the numbers from the issue report:

• prompt_tokens = 44341, cached_tokens = 41518
• input_cost_per_token = 1.4e-6, cache_read_input_token_cost = 2.6e-7

Before: 44341 * 1.4e-6 = $0.0621 (wrong)
After: 2823 * 1.4e-6 + 41518 * 2.6e-7 = $0.00395 + $0.01079 = $0.0147 (correct)

[greptile-apps]

Greptile Summary

This PR fixes Fireworks AI cost calculation so that cached tokens are billed at cache_read_input_token_cost (and cache-creation tokens at cache_creation_input_token_cost) rather than at the full input_cost_per_token rate. The change correctly extracts token buckets from usage.prompt_tokens_details and falls back to 0.0 when the cache cost fields are absent, leaving standard serverless pricing unaffected.

Confidence Score: 5/5

Safe to merge — the fix is correct and the only remaining finding is a minor defensive-programming suggestion.

The core logic is sound and consistent with how other providers (Dashscope) handle the same bucketing. The single P2 note about clamping non_cached_tokens at zero is a robustness improvement, not a blocker.

No files require special attention.

Important Files Changed

Filename	Overview
litellm/llms/fireworks_ai/cost_calculator.py	Correctly separates cached, cache-creation, and non-cached input tokens and applies their respective per-token rates; one minor concern about non_cached_tokens going negative with malformed usage data.

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart TD
    A[cost_per_token called] --> B{model in model_prices?}
    B -- yes --> C[get_model_info direct]
    B -- no --> D[get_base_model_for_pricing]
    D --> E[get_model_info for base model]
    C & E --> F{prompt_tokens_details present?}
    F -- yes --> G[extract cached_tokens\ncache_creation_tokens]
    F -- no --> H[cached_tokens = 0\ncache_creation_tokens = 0]
    G & H --> I[non_cached_tokens =\nprompt_tokens - cached - creation]
    I --> J[prompt_cost =\nnon_cached x input_cost\n+ cached x cache_read_cost\n+ creation x cache_creation_cost]
    J --> K[completion_cost =\ncompletion_tokens x output_cost]
    K --> L[return prompt_cost, completion_cost]

Loading

_{Reviews (1): Last reviewed commit: "fix: apply cache_read_input_token_cost t..." | Re-trigger Greptile}

[greptile-apps]

non_cached_tokens can go negative with malformed usage data

If the API ever returns cached_tokens + cache_creation_tokens > prompt_tokens (e.g. a rounding discrepancy or inconsistent response), non_cached_tokens is negative and prompt_cost ends up negative. Clamping to zero makes the cost calculation resilient to such inconsistencies.

Suggested change

	non_cached_tokens = usage["prompt_tokens"] - cached_tokens - cache_creation_tokens
	non_cached_tokens = max(0, usage["prompt_tokens"] - cached_tokens - cache_creation_tokens)

[codspeed-hq]

Merging this PR will not alter performance

✅ 16 untouched benchmarks

---

_{Comparing VANDRANKI:fix/fireworks-cache-cost (d0b62e7) with main (850fe59)}

[codecov]

Codecov Report

❌ Patch coverage is 0% with 9 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
litellm/llms/fireworks_ai/cost_calculator.py	0.00%	9 Missing ⚠️

📢 Thoughts on this report? Let us know!