Open-weight models,
sovereign UK infrastructure

Frontier reasoning, open weights

Alibaba's Qwen3.6 (qwen3_5_moe family) — vision, tool use, reasoning; 262K context.

Alibaba’s flagship multimodal family

OpenAI’s open-weight release

Google DeepMind’s efficient multimodal family

Purpose-built for code

Meta’s 70B workhorse

Alibaba's Qwen3-VL (qwen3_vl family) — vision, tool use, reasoning; 262K context.

Alibaba's Qwen3-VL Instruct (qwen3_vl family) — vision, tool use; 262K context.

Microsoft's Phi-4-reasoning (phi3 family) — reasoning; 32K context.

Microsoft's Phi-4-reasoning-plus (phi3 family) — reasoning; 32K context.

Alibaba's Qwen3-Coder-Next (qwen3_next family) — tool use; 262K context.

Z.ai's GLM-4.7-Flash (glm4_moe_lite family) — tool use, reasoning; 202K context.

Mistral AI's Ministral 3 (mistral3 family) — vision, tool use; 262K context.

Mistral AI's Ministral 3 Reasoning (mistral3 family) — tool use, reasoning; 262K context.

Mistral AI's Devstral Small (mistral3 family) — vision, tool use; 393K context.

Z.ai's GLM-OCR (glm_ocr family) — vision, OCR; 131K context.

Liquid AI's LFM2.5 (lfm2_moe family) — tool use; 128K context.

Liquid AI's LFM2.5 Thinking (lfm2 family) — tool use, reasoning; 128K context.

Liquid AI's LFM2.5-VL (lfm2_vl family) — vision; 128K context.

NVIDIA's Nemotron 3 Nano (nemotron family) — tool use, reasoning; 131K context.

Microsoft's Phi-4-mini-reasoning (phi3 family) — reasoning; 131K context.

IBM's Granite 4.0 Micro (granitemoehybrid family) — tool use; 131K context.

Mistral AI's Magistral Small (mistral family) — tool use, reasoning; 40K context.

Z.ai's GLM-4.6V-Flash (glm4v family) — vision; 131K context.

Google DeepMind's DiffusionGemma (diffusion_gemma family) — tool use, reasoning; 262K context.

Xiaomi's MiMo-V2.5 (mimo_v2 family) — vision, tool use, reasoning; 1048K context.

Xiaomi's MiMo-V2.5-Pro (mimo_v2 family) — tool use, reasoning; 1048K context.

BAAI's BGE Reranker v2 m3 (xlm-roberta family) — reranking; 8K context.

OpenAI Whisper speech-to-text, served in-process by the Pendra (whisper.cpp) backend.

FLUX.1 [schnell] fast 12B text-to-image, served in-process by the Pendra (stable-diffusion.cpp) backend.

Z-Image Turbo fast (few-step) text-to-image, served in-process by the Pendra (stable-diffusion.cpp) backend.

FLUX.2 klein 9B text-to-image, served in-process by the Pendra (stable-diffusion.cpp) backend.

SDXL Turbo fast (1-4 step) text-to-image, served in-process by the Pendra (stable-diffusion.cpp) backend.

FLUX.2 klein 4B fast text-to-image, served in-process by the Pendra (stable-diffusion.cpp) backend.

Qwen-Image 2512 high-fidelity text-to-image with strong text rendering, served in-process by the Pendra (stable-diffusion.cpp) backend.

Stable Diffusion XL 1.0 base text-to-image, served in-process by the Pendra (stable-diffusion.cpp) backend.

Stable Diffusion 3.5 Large Turbo fast text-to-image, served in-process by the Pendra (stable-diffusion.cpp) backend.

Stable Diffusion 1.5 text-to-image, served in-process by the Pendra (stable-diffusion.cpp) backend.

Nomic AI's nomic-embed-text-v1.5 (nomic_bert family) — embeddings; 2K context.

Alibaba's Qwen3-Embedding (qwen3 family) — embeddings; 40K context.

BAAI's BGE-M3 (xlm-roberta family) — embeddings; 8K context.

Google DeepMind's EmbeddingGemma (gemma3_text family) — embeddings; 2K context.