# Fireworks

HQ US · serves 14 tracked open-weights models. status page · privacy policy

models
14
verified
of overall top 10
8
cheapest on
vs best $
2.05×
ctx (median)
378k
tps (median)
115
ttft (median)
512ms
uptime
98.7%

# models

model rank $/M in $/M out $/M cache src
kimi-k3 #1 $3.00 $15.00 $0.30 openrouter
deepseek-v4-pro #2 $1.20 $1.20 $0.60 openrouter
glm-5.2 #3 $1.40 $4.40 $0.14 openrouter
kimi-k2.6 #4 $0.95 $4.00 $0.16 openrouter
minimax-m3 #6 $0.30 $1.20 hfrouter
kimi-k2.7-code #7 $0.95 $4.00 $0.19 openrouter
deepseek-v4-flash #8 $0.14 $0.28 litellm
glm-5.1 #10 $1.40 $4.40 $0.26 requesty
gpt-oss-120b #35 $0.15 $0.60 hfrouter
deepseek-v4-flash-0731 $0.22 $0.66 $0.01 openrouter
gpt-oss-20b $0.07 $0.30 $0.04 openrouter
inkling hfrouter
muse-glimmer-30b $0.35 $1.50 $0.04 openrouter
qwen3.8-2.4t-a95b hfrouter

# recent changes

when (UTC)modelchange
Sep 1 16:45 deepseek-v4-pro $/M input $1.74 → $1.20 ↓ 31.0%
Sep 1 16:45 deepseek-v4-pro $/M output $3.48 → $1.20 ↓ 65.5%
Aug 31 11:38 minimax-m2.7 no longer listed

all providers · machine-readable: providers.json