Skip to content

Self-hosted, Self-Improving, Federated Everything App — dev management, agent orchestration, persistent mind, digital twin, second brain, and creative pipeline from a single dashboard over Tailscale to port the user into the OS

feat: refresh local LLM catalog with Nemotron 3.5 Bolt 74B-A9B

[._.]/ Adam Eivy28bb7094ca2b
diff --git a/server/lib/localLlmCatalog.js b/server/lib/localLlmCatalog.js
index 9126207ba6..a9acb65168 100644
--- a/server/lib/localLlmCatalog.js
+++ b/server/lib/localLlmCatalog.js
@@ -991,6 +991,22 @@ export const LOCAL_LLM_CATALOG = [
ollama: 'qwen3-vl:30b',
lmstudio: 'lmstudio-community/Qwen3-VL-30B-A3B-Instruct-GGUF'
},
+ {
+ key: 'nemotron-3.5-bolt',
+ name: 'Nemotron 3.5 Bolt 74B-A9B',
+ category: 'vision',
+ recommendedFor: ['vision', 'reasoning', 'coding'],
+ params: '74B / 9B active',
+ size: '24 GB',
+ family: 'nemotron',
+ description: "NVIDIA's hybrid Mamba-2/Transformer vision-language MoE — multimodal reasoning, coding, and tool calling with 9B active parameters, compressed into a 24GB footprint for local execution.",
+ repository: 'nvidia/NVIDIA-Nemotron-3.5-Bolt-74B-A9B-24GB-GGUF',
+ capabilities: ['chat', 'reasoning', 'code', 'tools', 'vision'],
+ context: 262144,
+ ollama: 'nemotron-3.5-bolt',
+ lmstudio: 'lmstudio-community/NVIDIA-Nemotron-3.5-Bolt-74B-A9B-GGUF',
+ lmstudioAliases: ['nvidia/NVIDIA-Nemotron-3.5-Bolt-74B-A9B-24GB-GGUF']
+ },
{
key: 'llama3.2-vision-11b',
name: 'Llama 3.2 Vision 11B',
diff --git a/server/lib/localLlmCatalog.test.js b/server/lib/localLlmCatalog.test.js
index 13dac842e3..edf5b64700 100644
--- a/server/lib/localLlmCatalog.test.js
+++ b/server/lib/localLlmCatalog.test.js
@@ -217,6 +217,7 @@ describe('localLlmCatalog', () => {
expect(ollama.find((m) => m.key === 'qwen3.6-27b')?.id).toBe('qwen3.6:27b');
expect(ollama.find((m) => m.key === 'magistral-small-24b')?.id).toBe('magistral:24b');
expect(ollama.find((m) => m.key === 'nemotron-3.5-lightning')?.id).toBe('nemotron-3.5-lightning:30b');
+ expect(ollama.find((m) => m.key === 'nemotron-3.5-bolt')?.id).toBe('nemotron-3.5-bolt');
});
});
@@ -249,6 +250,8 @@ describe('localLlmCatalog', () => {
.toEqual({ targetId: 'lmstudio-community/Magistral-Small-2509-GGUF', exact: true });
expect(mapModelToBackend('ollama', 'nemotron-3.5-lightning:30b', 'lmstudio'))
.toEqual({ targetId: 'lmstudio-community/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-GGUF', exact: true });
+ expect(mapModelToBackend('ollama', 'nemotron-3.5-bolt', 'lmstudio'))
+ .toEqual({ targetId: 'lmstudio-community/NVIDIA-Nemotron-3.5-Bolt-74B-A9B-GGUF', exact: true });
});
it('still maps models retired from the suggested-install catalog', () => {
diff --git a/server/lib/localModelHeuristics.js b/server/lib/localModelHeuristics.js
index ba5c696686..6dbc343034 100644
--- a/server/lib/localModelHeuristics.js
+++ b/server/lib/localModelHeuristics.js
@@ -108,6 +108,7 @@ const VISION_RE = new RegExp([
'smolvlm', 'internvl', 'cogvlm', 'glm-?4v', 'phi-?3\\.5?-vision',
'phi-?4-multimodal', 'got-ocr', 'idefics', 'fuyu', 'paligemma',
'kosmos', 'nanollava',
+ 'nemotron-?3\\.5?-bolt',
].join('|'), 'i');
// CLI providers whose underlying model is vision-capable and whose CLI accepts
diff --git a/server/lib/localModelHeuristics.test.js b/server/lib/localModelHeuristics.test.js
index b00782400d..ae10e196f8 100644
--- a/server/lib/localModelHeuristics.test.js
+++ b/server/lib/localModelHeuristics.test.js
@@ -166,6 +166,7 @@ describe('localModelHeuristics', () => {
'gemma4:31b',
'qwen3-vl:8b',
'minicpm-v4.6:1b',
+ 'nemotron-3.5-bolt',
]) {
expect(isVisionModel(id), id).toBe(true);
}
diff --git a/server/lib/systemCapabilities.js b/server/lib/systemCapabilities.js
index cf2dec1ee6..a00794f663 100644
--- a/server/lib/systemCapabilities.js
+++ b/server/lib/systemCapabilities.js
@@ -60,6 +60,7 @@ const LOCAL_LLM_HARDWARE_REQUIREMENTS = Object.freeze({
'gemma4-31b': Object.freeze({ minMemoryGb: 64 }),
'nemotron-3-nano-30b-a3b': Object.freeze({ minMemoryGb: 32 }),
'nemotron-3.5-lightning': Object.freeze({ minMemoryGb: 32 }),
+ 'nemotron-3.5-bolt': Object.freeze({ minMemoryGb: 32 }),
'qwen3.6-27b': Object.freeze({ minMemoryGb: 32 }),
'qwen3.5-122b-a10b': Object.freeze({ minMemoryGb: 96 }),
'qwen3-coder-30b': Object.freeze({ minMemoryGb: 32 }),