Model Catalog¶

Last updated: 2026-06-28

The mining dashboard and node runtime use model identifiers when creating sessions and matching miner capacity. The catalog below maps dashboard model IDs to display names, runtime identifiers, and capacity rules.

How To Read The Catalog¶

ID

Use the numeric ID when a session, dashboard view, or runtime setting asks for a model number.

Runtime

Use the runtime identifier when configuring a worker image or LLM runtime.

Mode

Check whether the session is dedicated, ephemeral, or hybrid before assuming a model is selectable.

Capacity

Extended models require active advertised node capacity outside dedicated sessions.

Model Selection Flow¶

flowchart TB
  Session["Create session"] --> Model["Choose model identifier"]
  Model --> Mode{"Session mode"}
  Mode -->|"Dedicated"| Dedicated["All configured catalog models selectable"]
  Mode -->|"Ephemeral / hybrid"| Base["Base models 0-12 mostly selectable"]
  Mode -->|"Ephemeral / hybrid"| Extended["Extended models 13-77 require active node capacity"]
  Base --> Disabled["Large disabled IDs<br/>3, 16, 17, 34, 43, 55"]
  Extended --> Capacity["Dashboard reads NodeSpec.getModelMetadata(node.owner)"]
  Capacity --> Count["Model capacity count"]
  Count --> Selectable["Selectable when capacity > 0"]

Capacity Rules¶

Rule	Meaning
Model identifier	Numeric ID stored on sessions and used by miners to select the runtime model.
Base catalog	IDs `0-12`; the dashboard treats these as generally selectable, except ID `3` outside dedicated mode.
Extended catalog	IDs `13-77`; ephemeral/hybrid sessions require active node capacity.
Dedicated mode	Dashboard allows all configured catalog models for dedicated sessions.
Hard-disabled outside dedicated	IDs `3`, `16`, `17`, `34`, `43`, and `55` are disabled for non-dedicated modes in the dashboard.
Node capacity source	Dashboard calls `NodeSpec.getModelMetadata(node.owner)` and counts advertised `modelIndexes` for active nodes.

Catalog¶

Base catalog IDs 0-12

Default dashboard catalog for ordinary session selection. ID 3 is treated as large and is disabled outside dedicated mode.

Extended catalog IDs 13-77

Requires active advertised node capacity for ephemeral or hybrid sessions.

Dedicated-only large models 3, 16, 17, 34, 43, 55

Dashboard hard-disables these IDs outside dedicated sessions.

Embedding models IDs 56-64

Embedding-focused Ollama models; still capacity-gated outside dedicated mode.

Each row below is formatted as a model card. The ID is the value sessions and dashboards use; the runtime identifier is the string operators map to a worker/runtime.

ID	Display name	Runtime identifier	Engine/category	Session selection note
0	LLaVA 1.5 7B Q4	`llava-v1.5-7b-q4`	Llamafile	Base catalog model.
1	DeepSeek R1 Distill Llama 8B Q4_K_M	`DeepSeek-R1-Distill-Llama-8B-Q4_K_M`	Llamafile	Base catalog model.
2	Meta Llama 3.1 8B Instruct Q4_K_M	`Meta-Llama-3.1-8B-Instruct.Q4_K_M`	Llamafile	Base catalog model.
3	Qwen QwQ 32B Q4_K_M	`Qwen_QwQ-32B-Q4_K_M`	Llamafile	Large model; disabled outside dedicated mode in dashboard.
4	Mistral 7B Instruct v0.3 Q4_0	`Mistral-7B-Instruct-v0.3.Q4_0`	Llamafile	Base catalog model.
5	Granite 3.2 8B Instruct Q4_K_M	`granite-3.2-8b-instruct-Q4_K_M`	Llamafile	Base catalog model.
6	Qwen 2.5 7B Instruct 1M Q4_K_M	`Qwen2.5-7B-Instruct-1M-Q4_K_M`	Llamafile	Base catalog model.
7	Google Gemma 3 4B Q6_K	`google_gemma-3-4b-it-Q6_K`	Llamafile	Base catalog model.
8	Qwen Qwen3 4B Q8_0	`Qwen_Qwen3-4B-Q8_0`	Llamafile	Base catalog model.
9	Google Gemma 3 12B Q4_K_M	`google_gemma-3-12b-it-Q4_K_M`	Llamafile	Base catalog model.
10	DeepSeek R1 Distill Qwen 14B Q4_K_M	`DeepSeek-R1-Distill-Qwen-14B-Q4_K_M`	Llamafile	Base catalog model.
11	Qwen 2.5 Coder 14B Instruct Q4_K_M	`Qwen2.5-Coder-14B-Instruct-Q4_K_M`	Llamafile	Base catalog model.
12	Gemma 3 270M (Ollama)	`gemma3:270m-gpu`	Ollama	Small Ollama model; base selectable.
13	GPT OSS 20B (Ollama)	`gpt-oss:20b-gpu`	Ollama	Extended model; capacity-gated outside dedicated mode.
14	DeepSeek R1 8B (Ollama)	`deepseek-r1:8b-gpu`	Ollama	Extended model; capacity-gated outside dedicated mode.
15	Granite 4 3B (Ollama)	`granite4:3b-gpu`	Ollama	Extended model; capacity-gated outside dedicated mode.
16	GPT OSS 120B (Ollama)	`gpt-oss:120b-gpu`	Ollama	Large model; disabled outside dedicated mode in dashboard.
17	DeepSeek R1 70B (Ollama)	`deepseek-r1:70b-gpu`	Ollama	Large model; disabled outside dedicated mode in dashboard.
18	Gemma 3 4B (Ollama)	`gemma3:4b-gpu`	Ollama	Extended model; capacity-gated outside dedicated mode.
19	Gemma 3 12B (Ollama)	`gemma3:12b-gpu`	Ollama	Extended model; capacity-gated outside dedicated mode.
20	Ministral 3 8B (Ollama)	`ministral-3:8b-gpu`	Ollama	Extended model; capacity-gated outside dedicated mode.
21	DeepSeek R1 14B (Ollama)	`deepseek-r1:14b-gpu`	Ollama	Extended model; capacity-gated outside dedicated mode.
22	Gemma 3 27B (Ollama)	`gemma3:27b-gpu`	Ollama	Extended model; capacity-gated outside dedicated mode.
23	Qwen 3 VL 4B (Ollama)	`qwen3-vl:4b-gpu`	Ollama	Extended model; capacity-gated outside dedicated mode.
24	Qwen 3 VL 8B (Ollama)	`qwen3-vl:8b-gpu`	Ollama	Extended model; capacity-gated outside dedicated mode.
25	Ministral 3 14B (Ollama)	`ministral-3:14b-gpu`	Ollama	Extended model; capacity-gated outside dedicated mode.
26	Qwen 3 4B (Ollama)	`qwen3:4b-gpu`	Ollama	Extended model; capacity-gated outside dedicated mode.
27	Qwen 3 8B (Ollama)	`qwen3:8b-gpu`	Ollama	Extended model; capacity-gated outside dedicated mode.
28	Mistral 7B (Ollama)	`mistral:7b-gpu`	Ollama	Extended model; capacity-gated outside dedicated mode.
29	Phi 3 3.8B (Ollama)	`phi3:3.8b-gpu`	Ollama	Extended model; capacity-gated outside dedicated mode.
30	Phi 3 14B (Ollama)	`phi3:14b-gpu`	Ollama	Extended model; capacity-gated outside dedicated mode.
31	LLAMA 3.2 1B (Ollama)	`llama3.2:1b-gpu`	Ollama	Extended model; capacity-gated outside dedicated mode.
32	LLAMA 3.2 3B (Ollama)	`llama3.2:3b-gpu`	Ollama	Extended model; capacity-gated outside dedicated mode.
33	LLAMA 3.1 8B (Ollama)	`llama3.1:8b-gpu`	Ollama	Extended model; capacity-gated outside dedicated mode.
34	LLAMA 3.1 70B (Ollama)	`llama3.1:70b-gpu`	Ollama	Large model; disabled outside dedicated mode in dashboard.
35	Phi 4 14B (Ollama)	`phi4:14b-gpu`	Ollama	Extended model; capacity-gated outside dedicated mode.
36	Dolphin 3 8B (Ollama)	`dolphin3:8b-gpu`	Ollama	Extended model; capacity-gated outside dedicated mode.
37	TinyLLAMA 1.1B (Ollama)	`tinyllama1.1b-gpu`	Ollama	Extended model; capacity-gated outside dedicated mode.
38	Falcon 3 3B (Ollama)	`falcon3:3b-gpu`	Ollama	Extended model; capacity-gated outside dedicated mode.
39	Falcon 3 7B (Ollama)	`falcon3:7b-gpu`	Ollama	Extended model; capacity-gated outside dedicated mode.
40	Falcon 3 10B (Ollama)	`falcon3:10b-gpu`	Ollama	Extended model; capacity-gated outside dedicated mode.
41	DevStral 24B (Ollama)	`devstral:24b-gpu`	Ollama	Extended model; capacity-gated outside dedicated mode.
42	Qwen3 Coder 30B (Ollama)	`qwen3-coder:30b-gpu`	Ollama	Extended model; capacity-gated outside dedicated mode.
43	LLAMA 4 16x17B (Ollama)	`llama4:16x17b-gpu`	Ollama	Large model; disabled outside dedicated mode in dashboard.
44	LLaVA 7B (Ollama)	`llava:7b-gpu`	Ollama	Extended model; capacity-gated outside dedicated mode.
45	LLaVA 13B (Ollama)	`llava:13b-gpu`	Ollama	Extended model; capacity-gated outside dedicated mode.
46	LLaVA 34B (Ollama)	`llava:34b-gpu`	Ollama	Extended model; capacity-gated outside dedicated mode.
47	Nemotron 3 Nano 30B (Ollama)	`nemotron-3-nano:30b-gpu`	Ollama	Extended model; capacity-gated outside dedicated mode.
48	Olmo 3 7B (Ollama)	`olmo-3:7b-gpu`	Ollama	Extended model; capacity-gated outside dedicated mode.
49	Olmo 3 32B (Ollama)	`olmo-3:32b-gpu`	Ollama	Extended model; capacity-gated outside dedicated mode.
50	Olmo 2 7B (Ollama)	`olmo2:7b-gpu`	Ollama	Extended model; capacity-gated outside dedicated mode.
51	Olmo 2 13B (Ollama)	`olmo2:13b-gpu`	Ollama	Extended model; capacity-gated outside dedicated mode.
52	Orca Mini 3B (Ollama)	`orca-mini:3b-gpu`	Ollama	Extended model; capacity-gated outside dedicated mode.
53	Orca Mini 7B (Ollama)	`orca-mini:7b-gpu`	Ollama	Extended model; capacity-gated outside dedicated mode.
54	Orca Mini 13B (Ollama)	`orca-mini:13b-gpu`	Ollama	Extended model; capacity-gated outside dedicated mode.
55	Orca Mini 70B (Ollama)	`orca-mini:70b-gpu`	Ollama	Large model; disabled outside dedicated mode in dashboard.
56	Nomic Embed Text v1.5 (Embedding Model)	`nomic-embed-text:v1.5-gpu`	Embedding / Ollama	Extended model; capacity-gated outside dedicated mode.
57	MXBAI Embed Large 335M (Embedding Model)	`mxbai-embed-large:335m-gpu`	Embedding / Ollama	Extended model; capacity-gated outside dedicated mode.
58	BGE M3 567M (Embedding Model)	`bge-m3:567m-gpu`	Embedding / Ollama	Extended model; capacity-gated outside dedicated mode.
59	All MiniLM 22M (Embedding Model)	`all-minilm:22m-gpu`	Embedding / Ollama	Extended model; capacity-gated outside dedicated mode.
60	All MiniLM 33M (Embedding Model)	`all-minilm:33m-gpu`	Embedding / Ollama	Extended model; capacity-gated outside dedicated mode.
61	Embedding Gemma 300M (Embedding Model)	`embeddinggemma:300m-gpu`	Embedding / Ollama	Extended model; capacity-gated outside dedicated mode.
62	Snowflake Arctic Embed 335M (Embedding Model)	`snowflake-arctic-embed:335m-gpu`	Embedding / Ollama	Extended model; capacity-gated outside dedicated mode.
63	Snowflake Arctic Embed v2 568M (Embedding Model)	`snowflake-arctic-embed2:568m-gpu`	Embedding / Ollama	Extended model; capacity-gated outside dedicated mode.
64	Granite Embedding 278M (Embedding Model)	`granite-embedding:278m-gpu`	Embedding / Ollama	Extended model; capacity-gated outside dedicated mode.
65	GLM 4.7 Flash Q4_K_M (Ollama)	`glm-4.7-flash:q4_K_M-gpu`	Ollama	Extended model; capacity-gated outside dedicated mode.
66	LFM2 24B (Ollama)	`lfm2:24b-gpu`	Ollama	Extended model; capacity-gated outside dedicated mode.
67	LFM2.5 Thinking 1.2B (Ollama)	`lfm2.5-thinking:1.2b-gpu`	Ollama	Extended model; capacity-gated outside dedicated mode.
68	Qwen 3.5 0.8B (Ollama)	`qwen3.5:0.8b-gpu`	Ollama	Extended model; capacity-gated outside dedicated mode.
69	Qwen 3.5 2B (Ollama)	`qwen3.5:2b-gpu`	Ollama	Extended model; capacity-gated outside dedicated mode.
70	Qwen 3.5 4B (Ollama)	`qwen3.5:4b-gpu`	Ollama	Extended model; capacity-gated outside dedicated mode.
71	Qwen 3.5 9B (Ollama)	`qwen3.5:9b-gpu`	Ollama	Extended model; capacity-gated outside dedicated mode.
72	Qwen 3.5 27B (Ollama)	`qwen3.5:27b-gpu`	Ollama	Extended model; capacity-gated outside dedicated mode.
73	Gemma 4 E4B (Ollama)	`gemma4:e4b-gpu`	Ollama	Extended model; capacity-gated outside dedicated mode.
74	Gemma 4 26B (Ollama)	`gemma4:26b-gpu`	Ollama	Extended model; capacity-gated outside dedicated mode.
75	Gemma 4 31B (Ollama)	`gemma4:31b-gpu`	Ollama	Extended model; capacity-gated outside dedicated mode.
76	Qwen 3.6 27B (Ollama)	`qwen3.6:27b-gpu`	Ollama	Extended model; capacity-gated outside dedicated mode.
77	Qwen 3.6 35B (Ollama)	`qwen3.6:35b-gpu`	Ollama	Extended model; capacity-gated outside dedicated mode.

Operator Metadata¶

Area	What it shows
Dashboard model dialog	Static ID-to-name catalog shown to operators.
Dashboard session modal	Capacity rules, disabled model IDs, and `NodeSpec.getModelMetadata` capacity reads.
`LLMEngine.MODEL_ID_TO_INDEX`	Runtime mapping from model identifier string to numeric model ID.
`LLMDockerWorkerManager`	Container/image/port mapping for model workers, currently covering IDs `0-77`.
`tools/cognitivelevel/get_modelmetadata.py`	Operator tool for reading model metadata for a node address.

Availability Notes¶

This is the current catalog, not a guarantee that every model is live on every network.
Public availability depends on active node capacity, session mode, model disable rules, and deployment policy.
Customer-facing model lists include live router/session/model availability and last verified date.