Model catalogue · api.axforge.ai/v1
Open-weight models behind one OpenAI-compatible API, running on hardware AxForge owns in the EU: on the serverless API or deployed on dedicated hardware. Your prompts stay in the EU and are never kept or trained on: how we run the models, where data lives.
The newest
Fast chat and agents that read a whole codebase or a stack of contracts in one go: a 500,000-token context, it reads images, and it thinks when you ask. Served from Stockholm, on the same key as every other model.
| Model name | qwen3.8-flash-next on /v1/chat/completions |
|---|---|
| Context | 500,000 tokens |
| Input | $0.13 / 1M tokens |
| Cached input | $0.07 / 1M tokens |
| Output | $0.69 / 1M tokens |
Serverless Models
One key and one OpenAI-compatible API, api.axforge.ai/v1, with the region pinned on the key.
Each model's name goes in the model field of its endpoint.
The daily driver: assistants, tool calling, RAG. 262,144-token context.
qwen3.8-27b · /v1/chat/completions
$0.13 in · $0.69 out / 1M
Thinks before it answers: reasoning, code, agents. Text only.
glm-5.3-flash · /v1/chat/completions
$0.13 in · $0.69 out / 1M
Fast agents and tool use; thinks when you ask. Text only.
mimo-v2.6-flash · /v1/chat/completions
$0.13 in · $0.69 out / 1M
Yes or no, a choice or a score, with probabilities. Jev-compatible.
xev-latest · /v1/systemone
$0.025 / 1M in · output free
Vectors for semantic search and RAG.
qwen3-embed · /v1/embeddings
$0.009 / 1M
The passages that answer a question, first.
bge-reranker-v2-m3 · /v1/rerank
$0.01 / 1M
Posters, banners and UI shots with readable text.
ernie-image-turbo · /v1/images/generations
$0.06 / image
Photo edits from one sentence.
flux2-klein-4b · /v1/images/edits
$0.07 / edit
Speech to text: transcription and voice input.
whisper · /v1/audio/transcriptions
$0.006 / minute
Fast text to speech.
piper · /v1/audio/speech
$3.30 / 1M characters
Full songs and soundtracks from a prompt.
minimax-music3 · /v1/audio/music
$0.04 / 10 seconds
PDFs, scans and photos to text.
ocr-eu · /v1/ocr
$0.003 / page
Every price on the pricing page. Chat prices are launch pricing; list prices in EUR.
How to use them
1. Sign up and take your key. Your API key is made with your account, in the console. Every new account includes 3M free serverless tokens every 30 days.
2. Make the call. The quickstart in curl, Python or JavaScript, then each endpoint's page: chat, embeddings, rerank, Xev, images, speech and music, documents to text.
3. Or keep your tool. Connect Claude Code, Codex, Cline, OpenCode, LangChain, n8n and more, or take a Coder plan for a coding agent.
4. Stuck? The model list in the docs, the forum, or an engineer.
Managed GPU
Validated on AxForge hardware and deployed on a dedicated NVIDIA DGX Spark for your traffic only, as an OpenAI-compatible endpoint on your own machine, operated by AxForge. Hardware from $0.77/hour; managed service quoted per deployment.
| Model | Type | Availability | Best for | Price | |
|---|---|---|---|---|---|
| Qwen-Image | Image gen + edit | Available as managed deployment | Generation and editing on one dedicated machine | Quoted per deployment | Request deployment |
| Qwen3.6 35B A3B | LLM (MoE) | Available as managed deployment | High-throughput chat on a dedicated Spark | Quoted per deployment | Request deployment |
| Qwen3 30B A3B | LLM (MoE) | Available as managed deployment | Efficient volume workloads | Quoted per deployment | Request deployment |
| Gemma-4 26B | LLM (vision) | Available as managed deployment | Long documents (262,144-token context) plus vision | Quoted per deployment | Request deployment |
| Mistral Small 3.2 | LLM | Available as managed deployment | Compact European-family workhorse | Quoted per deployment | Request deployment |
| SDXL | Image generation | Available as managed deployment | High-volume light image generation | Quoted per deployment | Request deployment |
The open-source catalogue
Beyond what we serve today, AxForge tracks the open-weight models the labs keep shipping: DeepSeek, Qwen, Llama, Mistral, Gemma, FLUX, Wan and more. Each is available as a managed deployment on EU-hosted hardware, with the commercial licence arranged where one is required.
571 models
Chat, agents, tool use, reasoning and vision-language.
Browse all 571 language models →288 models
Generation and editing from text, and text- or image-to-video.
Browse 138 image → · 150 video →236 models
Speech-to-text, text-to-speech and music generation.
Browse all 236 audio & music →354 models
Embeddings, rerankers and the building blocks around a model: search, retrieval, classification.
Browse all 354 tools →Serverless or dedicated
Serverless Models · Available
The serverless API. One key, OpenAI-compatible endpoints for chat, embeddings, images and audio. Streaming, tool calls and usage reporting. Requests are pinned to an EU region on the key.
Managed Endpoint · Available
The same models and API, with your calls run ahead of shared traffic, at a fixed price by month or year. See Managed Endpoint.
Managed GPU · Request deployment
A model validated on AxForge hardware, deployed on a dedicated NVIDIA DGX Spark and operated for you, by the hour, week, month or year. An OpenAI-compatible endpoint on your own machine, serving only your traffic. See Managed GPU and the DGX Spark rental page.
Data & privacy
Prompts and completions are processed in memory in Sweden, not written to disk, not logged, not retained, and never used to train anything. We keep only request metadata (token counts, timestamps, status) for billing and operations. The full policy is at axforge.ai/privacy; the details are in the trust centre and the DPA.
FAQ
Qwen3.8 27B, GLM-5.3-Flash and
MiMo-V2.6-Flash (chat),
Qwen3.8 Flash Next (chat),
Xev (decisions),
Qwen3 Embedding,
ERNIE Image Turbo (image generation),
FLUX.2 Klein 4B (image editing),
MiniMax Music 3 (music generation),
Whisper (speech-to-text) and
Piper (text-to-speech). All
behind one OpenAI-compatible API at api.axforge.ai/v1.
The model is validated on AxForge hardware and deployed on a dedicated NVIDIA DGX Spark for your traffic only, with an OpenAI-compatible endpoint, rented by the hour, week, month or year. The managed service is quoted per deployment. Request deployment to start.
Serverless models run in Stockholm, Sweden (region eu-se-1),
on hardware AxForge owns and operates; dedicated GPU workloads run from
Málaga, Spain (eu-es-1). TLS terminates in the EU.
No. Prompts and completions are processed in memory in Sweden, not written to disk, not logged, not retained, never used to train. Only request metadata is kept for billing and operations. See the privacy policy.
Qwen3.8 27B is $0.13 per 1M input tokens and $0.69 per 1M output tokens (launch pricing). Embeddings are $0.009 per 1M tokens, images $0.06 each, edits $0.07, music $0.04 per 10 seconds, speech-to-text $0.006 per minute and text-to-speech $3.30 per 1M characters. Managed deployments are quoted per deployment. Request deployment.
