ONE API FOR AI AGENTS

Everything your AI agents need. One API.

Inference, realtime voice, and knowledge retrieval — powered by open models (Qwen, Llama, DeepSeek, Whisper, Flux) on a distributed GPU grid. One OpenAI-compatible endpoint for the whole agent stack.

Start free Browse models →

Run agents anywhere. Route inference to any available GPU. Powered by Grid

Your PCShare idle compute

Runs onRTX 30/40/50Apple Silicon

Group GPUsPower from group

Runs onDGX Spark

RTX WorkstationsPro power, shared

Runs onRTX A6000RTX PRO 6000

Community NodesStronger together

Runs onDGX Station

BUILD AI AGENTS

ReasoningLLM · Planning

VisionVLM · OCR

VoiceRealtime STT·LLM·TTS

Charbit VVoice Bridge

Knowledge / RAGEmbeddings · Rerank

GenerationImage · Video · Music

Tools & CodeFunctions · Sandbox

API

Your PCShare idle compute

Group GPUsPower from group

RTX WorkstationsPro power, shared

Community NodesStronger together

ReasoningLLM · Planning

VisionVLM · OCR

VoiceRealtime STT·LLM·TTS

Charbit VVoice Bridge

Knowledge / RAGEmbeddings · Rerank

GenerationImage · Video · Music

Tools & CodeFunctions · Sandbox

Every modality served from a distributed Grid GPU farm.

LLM

QwenLlamaDeepSeekGemma

Embedding

BGEGTEE5

Rerank

BGE-rerankerJina

STT

SenseVoiceWhisper

TTS

CosyVoiceVibeVoiceGPT-SoVITS

Image

Qwen-ImageFLUX

Video

MiniMax H3Wan 2.2LatentSync

Vision

Qwen-VLInternVL

DESKTOP CLIENT

Put your idle GPU on the Grid.

Install OpenAlchemy Engine Desktop and your idle GPU joins Grid — use the API for free, or earn credits every time someone else's job runs on your machine and spend them on API calls or in Charbit. Your call. The dashboard shows live GPU status, loaded models, queue, and logs.

OpenAlchemy Engine — v0.6.27

Dashboard — live GPU, CPU, memory, disk, and host

Worker — runtime slots and live request logs

Discover — browse and download open models

Dashboard — live GPU, CPU, memory, disk, and hostWorker — runtime slots and live request logsModels — everything cached on your diskDiscover — browse and download open models

Download for WindowsmacOS ・ Coming soon

OpenAlchemy Engine ・ v0.6.27

PLAYGROUND

Try every modality. One API.

LLMEmbeddingBetaRerankBetaSTTBetaTTSBetaImageBetaVideoBetaVisionBeta

LLM (Chat / Instruct)

Chat · Summarize · Code

Standard chat completions endpoint — for chat, summarization, code generation, and tool use.

Read docs →

# 1) discover an online chat model — Grid's set changes constantly,
#    don't hardcode an id in your code.
MODEL=$(curl -s https://api.openalchemy.io/v1/models \
  -H "Authorization: Bearer $OPENALCHEMY_API_KEY" \
  | jq -r '.data[] | select(.online and .endpoint_type=="chat") | .id' \
  | head -1)

# 2) chat
curl https://api.openalchemy.io/v1/chat/completions \
  -H "Authorization: Bearer $OPENALCHEMY_API_KEY" \
  -H "X-Project-Id: $YOUR_PROJECT_ID" \
  -H "Content-Type: application/json" \
  -d "{
    \"model\": \"$MODEL\",
    \"messages\": [
      {\"role\": \"user\", \"content\": \"Hello, OpenAlchemy.\"}
    ]
  }"

try it

Live requests run in Grid Playground with your own account and API key. Sign up is free.

Embedding

BetaSearch · Similarity · RAG

Turn text into vectors for semantic search, similarity, and RAG pipelines.

Read docs →

# 1) discover an online embedding model — Grid's set changes constantly,
#    don't hardcode an id in your code.
MODEL=$(curl -s https://api.openalchemy.io/v1/models \
  -H "Authorization: Bearer $OPENALCHEMY_API_KEY" \
  | jq -r '.data[] | select(.online and .endpoint_type=="embedding") | .id' \
  | head -1)

# 2) embed
curl https://api.openalchemy.io/v1/embeddings \
  -H "Authorization: Bearer $OPENALCHEMY_API_KEY" \
  -H "X-Project-Id: $YOUR_PROJECT_ID" \
  -H "Content-Type: application/json" \
  -d "{
    \"model\": \"$MODEL\",
    \"input\": \"Where is the best ramen in Tokyo?\"
  }"

try it

Live requests run in Grid Playground with your own account and API key. Sign up is free.

Reranker

BetaReorder · Relevance scoring

Reorder search results by relevance and lift top-K precision.

Read docs →

# 1) discover an online reranker — Grid's set changes constantly,
#    don't hardcode an id in your code.
MODEL=$(curl -s https://api.openalchemy.io/v1/models \
  -H "Authorization: Bearer $OPENALCHEMY_API_KEY" \
  | jq -r '.data[] | select(.online and .endpoint_type=="rerank") | .id' \
  | head -1)

# 2) rerank — results come back as {index, relevance_score},
#    index points into your documents array
curl https://api.openalchemy.io/v1/rerank \
  -H "Authorization: Bearer $OPENALCHEMY_API_KEY" \
  -H "X-Project-Id: $YOUR_PROJECT_ID" \
  -H "Content-Type: application/json" \
  -d "{
    \"model\": \"$MODEL\",
    \"query\": \"Where is the best ramen in Tokyo?\",
    \"documents\": [
      \"Ichiran has 24-hour shops in Shibuya.\",
      \"Tsuta in Sugamo holds a Michelin star.\",
      \"The Tsukiji market is famous for sushi.\",
      \"Afuri serves yuzu shio ramen in Ebisu.\"
    ],
    \"top_n\": 4
  }"

try it

Live requests run in Grid Playground with your own account and API key. Sign up is free.

STT (Speech → Text)

BetaTranscripts · Captions · Notes

Convert speech to text — for transcripts, captions, call analytics, and voice input.

Read docs →

# 1) discover an online STT model — Grid's set changes constantly,
#    don't hardcode an id in your code.
MODEL=$(curl -s https://api.openalchemy.io/v1/models \
  -H "Authorization: Bearer $OPENALCHEMY_API_KEY" \
  | jq -r '.data[] | select(.online and .endpoint_type=="stt") | .id' \
  | head -1)

# 2) transcribe (multipart upload)
curl https://api.openalchemy.io/v1/audio/transcriptions \
  -H "Authorization: Bearer $OPENALCHEMY_API_KEY" \
  -H "X-Project-Id: $YOUR_PROJECT_ID" \
  -F "model=$MODEL" \
  -F "file=@meeting.m4a" \
  -F "response_format=json"

try it

Live requests run in Grid Playground with your own account and API key. Sign up is free.

TTS (Text → Speech)

BetaNarration · Readout · Voice agents

Render text as natural-sounding audio — for narration, readouts, and voice agents.

Read docs →

# 1) discover an online TTS model — Grid's set changes constantly,
#    don't hardcode an id in your code.
MODEL=$(curl -s https://api.openalchemy.io/v1/models \
  -H "Authorization: Bearer $OPENALCHEMY_API_KEY" \
  | jq -r '.data[] | select(.online and .endpoint_type=="tts") | .id' \
  | head -1)

# 2) synthesize — returns WAV bytes. Built-in voices: GET /v1/audio/voices
curl https://api.openalchemy.io/v1/audio/speech \
  -H "Authorization: Bearer $OPENALCHEMY_API_KEY" \
  -H "Content-Type: application/json" \
  -d "{
    \"model\": \"$MODEL\",
    \"voice\": \"en_alice\",
    \"input\": \"OpenAlchemy turns idle GPUs into an open AI inference grid. Run a node on a 4090 or 5090 and earn credits, or call our APIs and pay per token. Open weights, distributed compute.\"
  }" \
  --output openalchemy_pitch.wav

try it

Live requests run in Grid Playground with your own account and API key. Sign up is free.

Image (gen / edit)

BetaGenerate · Edit

Generate images from text and edit them from reference images with Qwen-Image, Qwen-Image-Edit and FLUX 3 Klein, running on Grid GPUs.

Read docs →

curl https://api.openalchemy.io/v1/images/local/generations \
  -H "Authorization: Bearer $OPENALCHEMY_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "flux3-klein-local",
    "prompt": "A neon-lit Tokyo alley at dusk, photorealistic.",
    "size": "1024x1024",
    "steps": 16,
    "response_format": "b64_json"
  }' \
  | jq -r '.data[0].b64_json' | base64 -d > image.png

try it

Live requests run in Grid Playground with your own account and API key. Sign up is free.

qwen-image-2512
qwen-image-edit-2511
flux3-klein-local

Video (gen)

BetaText · Image · Reference · Audio to video, lip-sync

MiniMax H3 and Wan 2.2: shots with sound, characters that speak or sing to your audio, motion transfer, and lip-sync.

Read docs →

# 1) submit a shot — returns 202 with {id, status}
VIDEO_ID=$(curl -s https://api.openalchemy.io/v1/videos \
  -H "Authorization: Bearer $OPENALCHEMY_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "minimax-h3",
    "mode": "t2v",
    "prompt": "A drone shot flying over Mt. Fuji at sunrise.",
    "duration": 5,
    "resolution": "768P",
    "aspect_ratio": "16:9"
  }' | jq -r '.id')

# 2) poll until status is "succeeded" (queued / running / succeeded / failed)
curl -s https://api.openalchemy.io/v1/videos/$VIDEO_ID \
  -H "Authorization: Bearer $OPENALCHEMY_API_KEY" | jq -r '.status'

# 3) download the mp4
curl -o shot.mp4 https://api.openalchemy.io/v1/videos/$VIDEO_ID/content \
  -H "Authorization: Bearer $OPENALCHEMY_API_KEY"

try it

Live requests run in Grid Playground with your own account and API key. Sign up is free.

minimax-h3
wan2.2-s2v
wan2.2-i2v
wan2.2-animate
latentsync-1.6

Vision (VLM / OCR)

BetaImage understanding · OCR · Docs

Multimodal chat that accepts images, plus document OCR and table extraction.

Read docs →

# 1) discover an online vision-capable (VL) chat model — Grid's set
#    changes constantly, don't hardcode an id in your code.
MODEL=$(curl -s https://api.openalchemy.io/v1/models \
  -H "Authorization: Bearer $OPENALCHEMY_API_KEY" \
  | jq -r '.data[] | select(.online and .endpoint_type=="chat" and (.id | test("vl|vision"; "i"))) | .id' \
  | head -1)

# 2) ask about an image (https URL or data: URL)
curl https://api.openalchemy.io/v1/chat/completions \
  -H "Authorization: Bearer $OPENALCHEMY_API_KEY" \
  -H "X-Project-Id: $YOUR_PROJECT_ID" \
  -H "Content-Type: application/json" \
  -d "{
    \"model\": \"$MODEL\",
    \"messages\": [
      {\"role\": \"user\", \"content\": [
        {\"type\": \"text\", \"text\": \"What is in this image?\"},
        {\"type\": \"image_url\", \"image_url\": {\"url\": \"https://example.com/receipt.jpg\"}}
      ]}
    ]
  }"

try it

Live requests run in Grid Playground with your own account and API key. Sign up is free.

WHY OPENALCHEMY

Why an API, not local inference?

Anywhere

Your agent runs anywhere, not just on the GPU box

Local runtimes tie inference to one machine. The OpenAlchemy API frees your agent to live on a phone, a Lambda, a cron, a browser — call from anywhere over HTTPS. The GPU stays on Grid.

Cost

Idle GPUs, woken up

Pay for time on aggregated idle capacity from a global pool instead of holding dedicated cloud GPUs. Same throughput, lower unit cost.

Compatibility

Standard API, zero-cost migration

Familiar /v1/chat/completions, /v1/embeddings endpoints — existing SDKs work as-is. Swap the base URL and the API key — that's it.

FOUNDATION

A foundation that holds in production

Industry-standard REST API
Per-project billing and quotas
Per-domain rate limits and access control
Auditable request logs, visible only to project owners

START TODAY

Ready to turn inference cost into something closer to alchemy?

Get started free See pricing

Sign up for updates

Hear about new releases, features, and updates from the OpenAlchemy team.