Pick a base. Bring your data.
Every model below ships with curated training recipes, hardware defaults, and a private inference endpoint. Pricing is on the card — no quotes, no calls.
Instruction-tune frontier open models on your domain data.
Llama 3.3 70B
Meta · 128k ctx
Frontier open-weights generalist. Best-in-class for instruction-tuning at the 70B tier.
Qwen3 72B
Alibaba · 256k ctx
Strong multilingual generalist with long context. Excellent for non-English and code tasks.
Llama 3.1 8B
Meta · 128k ctx
Workhorse 8B. Cheap to tune, fast to serve, ships on one GPU.
Qwen3 8B
Alibaba · 256k ctx
8B with long context. Strong multilingual and code performance per dollar.
Teach vision-language models to read screens, charts, and docs.
Qwen2.5-VL 72B
Alibaba · Image · video · OCR
Frontier open VLM. Great for chart, table, and document understanding.
Qwen2.5-VL 7B
Alibaba · Image · video · OCR
7B VLM that fits on one GPU. Ideal for screen and document understanding agents.
Vision-language-action models for robotics and manipulation.
Pi-0
Physical Intelligence · Bimanual · 50Hz
Generalist VLA from Physical Intelligence. Bimanual manipulation out of the box.
OpenVLA 7B
Stanford / TRI · Single-arm · 30Hz
Open-source 7B VLA. Solid baseline for single-arm manipulation tasks.
LoRA and full fine-tunes for image and video generators.
FLUX.1 [dev]
Black Forest Labs · Up to 2048×2048
Frontier text-to-image. The default for character, product, and brand LoRAs.
FLUX.1 [schnell]
Black Forest Labs · 1–4 step inference
Distilled FLUX for cheap, fast generation. LoRA-only for license reasons.
SDXL 1.0
Stability AI · 1024×1024
Battle-tested SDXL. Cheapest tier when you need a custom diffusion model.