Skip to content
PricingSaaS
Sign in
Back
Pricing briefing
FFireworks
Fireworks
PACKAGING CHANGE

Fireworks added Kimi K3.1 Pro to its serverless rate table

By , Co-founder, PricingSaaS

Published

Fireworks added Kimi K3.1 Pro as a new model on its serverless rate table.

See before and after

Fireworks’ pricing page · Change detected

What changed

Fireworks added Kimi K3.1 Pro as a new row in its Highlighted Serverless rates table. Per Fireworks's pricing page as of Nov 14, 2026, the model is listed at $2.00 per 1M input tokens, $0.20 per 1M cached input tokens, and $10.00 per 1M output tokens on the standard tier. A separately priced Priority tier for the same model lists at $3.00, $0.30, and $15.00 per 1M tokens for input, cached input, and output respectively, according to the same Nov 14, 2026 page.

Why it matters

For product leaders watching inference pricing, this is less about a single model's price than about where Fireworks keeps putting new models: directly into the Highlighted Serverless table rather than a separate or niche listing. Pairing that with a separately priced Priority tier for the same model shows Fireworks setting a standard price alongside a higher, latency-priority price for new releases as a matter of course, not as an exception.

Teams evaluating serverless inference providers should treat that Priority/standard split as a lever to negotiate or avoid rather than a fixed line item, and watch whether Fireworks keeps extending it to future models at the same cadence.

In context

Fireworks has made 22 real pricing-page changes since tracking began on Nov 17, 2024, averaging one every 32 days per Pulse's record; its previous change, 19 days earlier on Oct 26, 2026, added SOC 2 and ISO compliance badges [[fireworks/2026W44]]. Model additions are Fireworks's single most common type of change — nine tracked, most recently new training-API models on Oct 17, 2026 [[fireworks/2026D290]] — so this fits an established pattern, not a new strategy. Modal added its own new model, Inkling 2 Small, to its rate card a day earlier, on Nov 13, 2026 [[modal/2026D317]], a reminder that new-model listings are landing across serverless providers on a near-weekly cadence this month. Pulse tracked 738 feature-addition changes across 549 companies in the last 180 days — 23.9% of all 3,083 changes tracked — so a listing like this now reads as routine rather than notable; the sharper read is which provider gets to a new model first and at what price, not whether a provider adds it.

Common questions

How would I estimate my own cost on this model?

Multiply the per-1M-token rate by your expected input, cached-input, and output volume separately, and confirm which tier — standard or Priority — your workload actually needs before assuming the lower price applies.

Does Fireworks explain why Priority access costs more?

No — Fireworks's pricing page carries no stated reason for the Priority tier's higher price, and this briefing does not guess at one.

Sources