Astra Compute/Deploy

Deploy a machine

Pick a model, choose compute, configure the endpoint, settle payment, launch.

Llama 3.1 8B Instruct
Meta
Context
128K
Suggested GPU
RTX 4090 24GB
Est. VRAM
18 GB

Fast general chat and agent loops at low cost

Fallback available
Llama 3.1 70B Instruct
Meta
Context
128K
Suggested GPU
2× A100 80GB
Est. VRAM
140 GB

High-quality reasoning, production assistants

Fallback available
Mistral 7B Instruct
Mistral
Context
32K
Suggested GPU
RTX 3090 24GB
Est. VRAM
16 GB

Cheap, snappy completions and classification

Fallback available
Mixtral 8x7B Instruct
Mistral
Context
32K
Suggested GPU
2× RTX 4090 / A100 80GB
Est. VRAM
90 GB

MoE quality at dense-7B latency

Fallback available
Qwen 2.5 14B Instruct
Alibaba
Context
128K
Suggested GPU
L40S 48GB
Est. VRAM
30 GB

Multilingual chat and structured output

Fallback available
DeepSeek Coder V2 Lite
DeepSeek
Context
125K
Suggested GPU
RTX 4090 24GB
Est. VRAM
20 GB

Code completion, refactoring, repo Q&A

Fallback available
Custom model
Custom
Context
model-defined
Suggested GPU
Depends on weights
Est. VRAM
varies

Bring any model repository ID

No fallback