cameo · x86_64 · amdgpu
any AMD card.
it'll serve.
An Arch-based OS that turns any AMD GPU into an OpenAI-compatible endpoint. One old Radeon or a nine-card cluster — same two commands.
forge-01
0.0.0.0:9090
0tok / sec
RX 7900 XTX · 24GB
live
T1RX 7900 XTXqwen2.5-32b148
T2RX 6800llama-3.1-8b96
T3RX 580vulkan · idle—
01one card → a cluster
02what your card does
1train + serve7900 XTX · MI210
2serveRX 6800 · 6700 XT
3vulkanRX 580 · APUs
03detect · pull · serve
$ cameo detect 3 AMD GPUs · tiers 1/2/3 · vulkan OK $ cameo serve qwen2.5-32b --node forge-01 http://forge-01:9090/v1/chat/completions
04the fleet
fleet map · forge2 nodes · placement live
nownodes
nextthe map
v2clusters