Supported Models
Fleek provides optimized inference for leading open-source AI models across text, image, video, and 3D generation. All models run at $0.0025 per GPU-second with no per-token or per-prediction markup.
Large Language Models (LLMs)
| Model | Provider | Parameters | Context | Use Case |
|---|---|---|---|---|
| DeepSeek R1 | DeepSeek | 671B MoE (37B active) | 256K | Advanced reasoning, math, code |
| Kimi K2.5 | Moonshot AI | 1T MoE (32B active) | 256K | Multimodal, agentic workflows |
| GLM 4.7 | Z.ai | 355B MoE (32B active) | 200K | Code generation (73.8% SWE-bench) |
| Qwen3-235B | Alibaba | 235B | 128K | Multilingual, general reasoning |
| Llama 70B | Meta | 70B | 128K | Production workloads |
| gpt-oss-120b | Community | 120B | 32K | High throughput, clean IP |
Coding-Optimized
| Model | Provider | Parameters | Context | Highlights |
|---|---|---|---|---|
| Qwen3 Coder 480B | Alibaba | 480B MoE | 256K | 67% SWE-bench, Apache 2.0 |
| Qwen3 Coder 30B A3B | Alibaba | 30B MoE (3B active) | 256K | Lightweight, fast inference |
Image Generation
| Model | Provider | Quality | Resolution | Features |
|---|---|---|---|---|
| FLUX.2 | Black Forest Labs | Premium | 2048×2048 | Multi-reference editing, color control |
| Z-Image | Alibaba | Quality | 2048×2048 | DiT architecture, LoRA-friendly |
| Qwen Image 2512 | Alibaba | Quality | 2048×2048 | Prompt enhancement, style transfer |
| SDXL Turbo | Stability AI | Fast | 1024×1024 | Real-time, 1-step generation |
Image Editing
| Model | Provider | Features |
|---|---|---|
| Qwen Edit 2511 | Alibaba | Instruction-based editing, style transfer, object removal |
Video Generation
| Model | Provider | Duration | Resolution | Features |
|---|---|---|---|---|
| LTX-2 19B | Lightricks | Up to 20s | 4K | Audio sync, lip sync, 50fps |
| Wan Move | Alibaba | Up to 15s | 1080p | Image-to-video, motion control |
| Wan 2.2 14B | Alibaba | Up to 16s | 1080p | Text-to-video, temporal consistency |
| SeedVR | ByteDance | Up to 10s | 4K | Video upscaling, detail enhancement |
| Seedance 1.5 | ByteDance | Up to 12s | 1080p | Dance generation, beat sync |
3D Generation
| Model | Provider | Output Formats | Features |
|---|---|---|---|
| Hunyuan 3D-2 | Tencent | GLB, OBJ, FBX | Text-to-3D, image-to-3D, game-ready |
| Trellis 2 4B | Microsoft | GLB, USDZ, OBJ | Fast, MIT licensed, Apple support |
Model Selection Guide
For reasoning & analysis
→ DeepSeek R1 — Best-in-class chain-of-thought reasoning
For coding tasks
→ GLM 4.7 — Highest SWE-bench score (73.8%) → Qwen3 Coder 480B — Handles entire codebases
For multimodal workflows
→ Kimi K2.5 — Native image/screenshot analysis
For production scale
→ Llama 70B — Proven reliability, broad support
For image generation
→ FLUX.2 — Photorealistic quality → SDXL Turbo — Real-time generation
For video
→ LTX-2 19B — Only model with native audio sync
Coming Soon
We're constantly adding new models. Request a model on Discord or contact us.
Questions?
- Browse models: fleek.sh/models
- Pricing details: fleek.sh/docs/pricing
- Discord: discord.gg/h6K2cJU4np