# Scaleway

serves 12 tracked open-weights models.

models
12
verified
of overall top 10
1
cheapest on
1 models
vs best $
3.76×
ctx (median)
164k
tps (median)
92
ttft (median)
426ms

# models

model rank $/M in $/M out $/M cache src
glm-5.2 #3 $2.05 $6.27 hfrouter
qwen3.6-35b-a3b #16 $0.28 $1.71 hfrouter
qwen3.5-397b-a17b #17 $0.68 $4.10 hfrouter
gemma-4-26b-a4b-it #27 $0.28 $0.57 hfrouter
gpt-oss-120b #35 $0.17 $0.68 hfrouter
qwen3-235b-a22b-instruct-2507 #39 $0.85 $2.56 hfrouter
qwen3-coder-30b-a3b #43 $0.23 $0.91 hfrouter
gemma-3-27b-it #54 $0.25 $0.50 litellm
deepseek-v4-flash-0731 $0.46 $0.93 $0.09 requesty
devstral-2-2512 $0.40 $2.00 litellm
llama-3.3-70b-instruct $1.03 $1.03 hfrouter
mistral-small-3.2-24b $0.15 $0.35 litellm

# recent changes

when (UTC)modelchange
Sep 3 21:08 gpt-oss-120b $/M output $0.70 → $0.68 ↓ 2.3% [source switch]
Sep 3 16:33 gpt-oss-120b $/M output $0.68 → $0.70 ↑ 2.3% [source switch]
Sep 1 11:49 qwen3.5-397b-a17b $/M input $0.60 → $0.68 ↑ 14.0% [source switch]
Sep 1 11:49 qwen3.5-397b-a17b $/M output $3.60 → $4.10 ↑ 14.0% [source switch]
Sep 1 10:11 qwen3.5-397b-a17b $/M input $0.68 → $0.60 ↓ 12.3% [source switch]
Sep 1 10:11 qwen3.5-397b-a17b $/M output $4.10 → $3.60 ↓ 12.3% [source switch]

all providers · machine-readable: providers.json