VERDANTBLOOM

MODELS

Every model here is published by its original maker under a licence that allows commercial use. Generated from what the gateway serves: if it is listed, you can call it with "model": "<id>".

25 MODELS

MODELS
IDMODELPUBLISHERLICENCEINPUTSIZECONTEXTCOLD STARTPUBLISHER'S CARDGPUPRICE WHILE RUNNING
mistral-nemo-12bMistral Nemo Instruct 2407Mistral AIApache-2.0Text12B32kabout 44 s1 x 24 GB GPUfrom $0.22 / hmistralai/Mistral-Nemo-Instruct-2407
mistral-small-32-24bMistral Small 3.2 24B InstructMistral AIApache-2.0Text24B64kabout 145 s1 x 48 GB GPU (A40 class)from $0.46 / hmistralai/Mistral-Small-3.2-24B-Instruct-2506
qwen25vl-7bQwen2.5-VL 7B InstructQwen (Alibaba Cloud)Apache-2.0Text + images7B32kabout 82 s1 x 48 GB GPU (A40 class)from $0.46 / hQwen/Qwen2.5-VL-7B-Instruct
qwen3vl-2bQwen3-VL 2B InstructQwen (Alibaba Cloud)Apache-2.0Text + images2B32kabout 62 s1 x 24 GB GPUfrom $0.22 / hQwen/Qwen3-VL-2B-Instruct
qwen3vl-8bQwen3-VL 8B InstructQwen (Alibaba Cloud)Apache-2.0Text + images8B32kabout 38 s1 x 48 GB GPU (A40 class)from $0.46 / hQwen/Qwen3-VL-8B-Instruct
qwen38-27b-baseQwen3.8 27BQwen (Alibaba Cloud)Apache-2.0Text27B256kabout 64 s1 x 48 GB GPU (A40 class)from $0.46 / hQwen/Qwen3.8-27B
qwen36-35b-a3bQwen3.6 35B-A3BQwen (Alibaba Cloud)Apache-2.0Text35B MoE (3B active)256kabout 50 s1 x 48 GB GPU (A40 class)from $0.46 / hQwen/Qwen3.6-35B-A3B
qwen35-9bQwen3.5 9BQwen (Alibaba Cloud)Apache-2.0Text9B128kabout 55 s1 x 24 GB GPUfrom $0.22 / hQwen/Qwen3.5-9B
qwen25-14bQwen2.5 14B InstructQwen (Alibaba Cloud)Apache-2.0Text14B32kabout 176 s1 x 24 GB GPUfrom $0.22 / hQwen/Qwen2.5-14B-Instruct
gemma4-31bGemma 4 31B itGoogleApache-2.0Text + images31B64kabout 74 s1 x 48 GB GPU (A40 class)from $0.46 / hgoogle/gemma-4-31B-it
gemma4-26b-a4bGemma 4 26B-A4B itGoogleApache-2.0Text + images26B MoE (4B active)32kabout 51 s1 x 48 GB GPU (A40 class)from $0.46 / hgoogle/gemma-4-26B-A4B-it
gemma4-e4bGemma 4 E4B itGoogleApache-2.0TextE4B64kabout 95 s1 x 24 GB GPUfrom $0.22 / hgoogle/gemma-4-E4B-it
llama31-8bLlama 3.1 8B InstructMetaLlama 3.1 Community LicenceText8B32kabout 69 s1 x 24 GB GPUfrom $0.22 / hmeta-llama/Llama-3.1-8B-Instruct
gemma4-12bGemma 4 12B InstructGoogleApache-2.0Text + images12B32kabout 106 s1 x 24 GB GPUfrom $0.22 / hgoogle/gemma-4-12B-it
laguna-xsLaguna XS 2.1poolsideOpenMDW-1.1Text33B64kabout 67 s1 x 48 GB GPU (A40 class)from $0.46 / hpoolside/Laguna-XS-2.1
nemotron3-30b-a3bNemotron 3 Nano 30B-A3BNVIDIANVIDIA Open Model LicenseText32B (3B active)32kabout 62 s1 x 48 GB GPU (A40 class)from $0.46 / hnvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-BF16
deepseek-r1-qwen-7bDeepSeek-R1-Distill-Qwen-7BDeepSeekMITText7B128kabout 242 s1 x 24 GB GPUfrom $0.22 / hdeepseek-ai/DeepSeek-R1-Distill-Qwen-7B
deepseek-r1-qwen-14bDeepSeek-R1-Distill-Qwen-14BDeepSeekMITText14B64kabout 64 s1 x 48 GB GPU (A40 class)from $0.46 / hdeepseek-ai/DeepSeek-R1-Distill-Qwen-14B
deepseek-r1-qwen-32bDeepSeek-R1-Distill-Qwen-32BDeepSeekMITText32B32kabout 55 s1 x 48 GB GPU (A40 class)from $0.46 / hdeepseek-ai/DeepSeek-R1-Distill-Qwen-32B
deepseek-r1-llama-8bDeepSeek-R1-Distill-Llama-8BDeepSeekMIT (base model: Llama 3.1 Community License)Text8B32kabout 48 s1 x 24 GB GPUfrom $0.22 / hdeepseek-ai/DeepSeek-R1-Distill-Llama-8B
deepseek-r1-0528-qwen3-8bDeepSeek-R1-0528-Qwen3-8BDeepSeekMITText8B32kabout 40 s1 x 24 GB GPUfrom $0.22 / hdeepseek-ai/DeepSeek-R1-0528-Qwen3-8B
phi-4Phi-4MicrosoftMITText14B16kabout 39 s1 x 48 GB GPU (A40 class)from $0.46 / hmicrosoft/phi-4
phi-4-miniPhi-4-mini-instructMicrosoftMITText3.8B64kabout 48 s1 x 24 GB GPUfrom $0.22 / hmicrosoft/Phi-4-mini-instruct
phi-4-mini-reasoningPhi-4-mini-reasoningMicrosoftMITText3.8B64kabout 44 s1 x 24 GB GPUfrom $0.22 / hmicrosoft/Phi-4-mini-reasoning
gpt-oss-20bgpt-oss-20bOpenAIApache-2.0Text21B MoE (3.6B active)128kabout 118 s1 x 24 GB GPUfrom $0.22 / hopenai/gpt-oss-20b

CONTEXT is the model's window here; your level caps each request (see pricing). COLD START is the measured median time to the first token when the model has to start; once running it stays up 10 minutes after the last request. Quantised weights are standard quantisations of the publisher's release; the publisher's licence applies to your use of each model.