catalog

Model mappings

These ids map relay requests to a configured OpenAI-compatible provider. Catalog inclusion is not a claim of availability; effective status is shown on every card and by GET /v1/models.

active route: not configuredchat requests currently return 503api reference →

Llama 3.3 70B Instruct

llama-3.3-70b-instruct

unavailable

Open-weights flagship and the sensible default. Strong instruction following across eight languages with tool calling.

provider: Meta
context: 128k
speed: balanced

chattoolsopen-weightscatalog mapping

Llama 3.1 8B Instruct

llama-3.1-8b-instruct

unavailable

The workhorse small model. Cheap, quick and reliable for classification, extraction and routing.

provider: Meta
context: 128k
speed: fast

chatfastopen-weightscatalog mapping

Qwen2.5 72B Instruct

qwen-2.5-72b-instruct

unavailable

Excellent multilingual coverage and structured output. A strong pick for agentic workflows.

provider: Alibaba
context: 32k
speed: deep

chattoolsopen-weightscatalog mapping

Qwen2.5 Coder 32B

qwen-2.5-coder-32b

unavailable

Repository-scale code generation, diff editing and test writing. Currently the best open code model per parameter.

provider: Alibaba
context: 32k
speed: balanced

codeopen-weightscatalog mapping

Mistral Small 3.1

mistral-small-3.1

unavailable

Efficient multimodal model with vision support and a very low cost per token. Good latency-to-quality ratio.

provider: Mistral AI
context: 94k
speed: fast

chatvisionopen-weightscatalog mapping

Mistral Nemo 12B

mistral-nemo

unavailable

Compact long-context model built with NVIDIA. Handles 128k context comfortably for summarisation tasks.

provider: Mistral AI
context: 128k
speed: fast

chatfastopen-weightscatalog mapping

DeepSeek R1 Distill 70B

deepseek-r1-distill-70b

unavailable

Reasoning distilled from R1 into Llama 70B. Emits an explicit chain of thought before answering.

provider: DeepSeek
context: 128k
speed: deep

reasoningcodeopen-weightscatalog mapping

Gemma 3 27B IT

gemma-3-27b-it

unavailable

Compact multimodal model tuned for helpful, grounded responses with image understanding.

provider: Google
context: 94k
speed: balanced

chatvisionopen-weightscatalog mapping

Gemma 2 9B IT

gemma-2-9b-it

unavailable

Very small footprint, surprisingly good writing quality. Good for high-volume, low-stakes text.

provider: Google
context: 8k
speed: fast

chatfastopen-weightscatalog mapping

Phi-4 14B

phi-4

unavailable

Synthetic-data trained reasoning model that punches well above its size on maths and logic.

provider: Microsoft
context: 16k
speed: balanced

reasoningcodeopen-weightscatalog mapping