Run serious coding models.
Hosted API or your own GPU.

Ollama-simple. Tiered pricing, prefix-cache discounts, 262K context.

curl -fsSL https://oaica.com/install.sh | bash

Hosted API

OpenAI- and Anthropic-compatible. 262K context, vision, prefix-cache discounts (~55% hit rate on agent workloads).

curl -X POST https://api.oaica.com/v1/chat/completions \
  -H "Authorization: Bearer $OAICA_API_KEY" \
  -d '{"model":"oaica-35b-a3b-vision","messages":[{"role":"user","content":"Hello"}]}'

Self-host

Free. Pull models and run on your own infrastructure with oaica pull and oaica serve.

oaica pull oaica-nemotron-30b-a3b
oaica serve oaica-nemotron-30b-a3b

Claude Code ready

Context window auto-configured. Launch Claude Code directly on our models.

oaica launch claude --model oaica-35b-a3b-vision

4×

A100 fleet

55%

Prefix-cache hit rate

262K

Context length

Token

Usage-metered billing

Pricing

Per M tokens. Context-tiered. Cache-hit discounts for agent sessions.

Prompt size	Input (uncached)	Cache-hit	Output
≤ 32K	$0.05	$0.008	$0.28
32K–128K	$0.06	$0.008	$0.28
> 128K	$0.10	$0.008	$0.28

The bracket is chosen by your request's real prompt size. Cached prefix always bills at $0.008. Agent sessions that reuse context pay mostly cache rates.

Model	Input $/M	Output $/M
DeepSeek V4 Flash	$0.22–0.44	$0.66–1.32
KAT-Coder-Air	$0.15	$0.60
OAICA	$0.05–0.10	$0.28

Flat monthly plans available — talk to us at oaica@sprapp.com. Priority-tier keys skip large-context queuing.

Models

Generated from the gateway. Live at https://api.oaica.com/v1/models.

oaica-35b-a3b-vision Live

262,144 context · 32K max output · Text + image (2 images/request)

Tiered pricing above. Vision-capable for coding assistance with screenshots.

# OpenAI-compatible
curl -X POST https://api.oaica.com/v1/chat/completions \
  -H "Authorization: Bearer $OAICA_API_KEY" \
  -d '{"model":"oaica-35b-a3b-vision","messages":[{"role":"user","content":"Hello"}]}'

# CLI
oaica run oaica-35b-a3b-vision

# Claude Code
oaica launch claude --model oaica-35b-a3b-vision

oaica-nemotron-30b-a3b On-demand

262,144 context · Text-only · Reasoning + tool-calling

Currently on-demand (pool paused). Auto-configures context window for Claude Code.

# Self-host
oaica pull oaica-nemotron-30b-a3b
oaica serve oaica-nemotron-30b-a3b

# Check catalog
curl https://api.oaica.com/v1/catalog

Quickstart

Three paths. Pick yours.

1. Hosted API

curl -fsSL https://oaica.com/install.sh | bash
oaica signin
oaica run oaica-35b-a3b-vision
oaica launch claude --model oaica-35b-a3b-vision

2. Direct API

export OAICA_API_KEY=

curl -X POST https://api.oaica.com/v1/chat/completions \
  -H "Authorization: Bearer $OAICA_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"oaica-35b-a3b-vision","messages":[{"role":"user","content":"Hello"}],"stream":true}'

# Note: scripted clients must set a custom User-Agent
# (default Python-urllib is bot-blocked at the edge)
curl -X POST https://api.oaica.com/v1/chat/completions \
  -H "User-Agent: my-app/1.0" \
  -H "Authorization: Bearer $OAICA_API_KEY" \
  -d '{"model":"oaica-35b-a3b-vision","messages":[{"role":"user","content":"Hello"}]}'

3. Self-host

# List available models
curl https://api.oaica.com/v1/catalog

# Pull a model (25 GB for nemotron)
oaica pull oaica-nemotron-30b-a3b

# Serve it
oaica serve oaica-nemotron-30b-a3b

# Licensed models require OAICA_LICENSE_KEY
export OAICA_LICENSE_KEY=
oaica pull 
oaica serve

Run serious coding models.Hosted API or your own GPU.