Enable Fireworks Priority tier with live timing comparison - #56
Merged
Merged
Conversation
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Send all Fireworks agent inference requests with service_tier=priority to reduce overload rejection risk. Preserve existing model selection, reasoning settings, cache affinity, retries, and deadlines. Update model-cost accounting to published Priority prices and include the requested tier in model-attempt diagnostics. Workers AI settings and pricing remain unchanged.
The live comparison was run as a local experiment. Benchmark scripts and raw results are excluded from this PR; measured results are summarized below.
Live validation
Both production models accepted Priority requests. All 30 requests returned HTTP 200 and schema-valid responses, including all 15 Priority calls. The response did not echo the tier, so this establishes request acceptance rather than independent scheduling or billing confirmation.
Five samples per tier and workload, alternating order, no retries. No timeout or overload occurred, so no reliability improvement was established. Two Standard vision responses failed an exact color-list check; those were not API failures. Synthetic images are simpler than production storyboards, output lengths varied, and the benchmark ran locally rather than in the Worker. Priority was faster for GLM in this sample and slower for DeepSeek. It is not a latency guarantee.
Checks
Not deployed. The configured GLM and DeepSeek models have a 25% Priority premium. The optional GPT-OSS profile has a 20% premium, reflected in accounting.