harn-vm 0.8.128

Async bytecode virtual machine for the Harn programming language
Documentation 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211

# NVIDIA NIM hosted LLM routes. The API Catalog exposes 100+ OpenAI-compatible
# model IDs; Harn catalogues the current agent/coding routes that are useful as
# first-class selectors and lets the live refresh workflow surface the rest.

[models."nvidia/nemotron-3-ultra-550b-a55b"]
name = "Nemotron 3 Ultra 550B A55B (NVIDIA NIM)"
provider = "nvidia"
context_window = 1048576
logical_model = "nemotron-3-ultra-550b-a55b"
equivalence_group = "nemotron-3-ultra-550b-a55b"
served_variant = "nvidia-nim"
api_dialect = "openai_chat"
capabilities = ["tools", "streaming", "thinking"]
architecture = { parameter_count_b = 550.0, active_parameter_count_b = 55.0, moe = true, source_url = "https://build.nvidia.com/nvidia/nemotron-3-ultra-550b-a55b/modelcard", last_verified = "2026-06-20" }
tier = "frontier"
open_weight = true
strengths = ["agentic", "reasoning", "coding", "tool_use", "long_context"]
complementary_with = ["anthropic-claude", "openai-gpt", "google-gemini", "qwen", "deepseek", "kimi"]

[models."nvidia/nemotron-3-super-120b-a12b"]
name = "Nemotron 3 Super 120B A12B (NVIDIA NIM)"
provider = "nvidia"
context_window = 1048576
logical_model = "nemotron-3-super-120b-a12b"
equivalence_group = "nemotron-3-super-120b-a12b"
served_variant = "nvidia-nim"
api_dialect = "openai_chat"
capabilities = ["tools", "streaming", "thinking"]
architecture = { parameter_count_b = 120.0, active_parameter_count_b = 12.0, moe = true, source_url = "https://build.nvidia.com/nvidia/nemotron-3-super-120b-a12b/modelcard", last_verified = "2026-06-20" }
tier = "frontier"
open_weight = true
strengths = ["agentic", "reasoning", "coding", "tool_use", "long_context", "speed"]
complementary_with = ["anthropic-claude", "openai-gpt", "google-gemini", "qwen", "deepseek", "kimi"]

[models."nvidia/nemotron-3-nano-30b-a3b"]
name = "Nemotron 3 Nano 30B A3B (NVIDIA NIM)"
provider = "nvidia"
context_window = 262144
logical_model = "nemotron-3-nano-30b-a3b"
equivalence_group = "nemotron-3-nano-30b-a3b"
served_variant = "nvidia-nim"
api_dialect = "openai_chat"
capabilities = ["tools", "streaming", "thinking"]
architecture = { parameter_count_b = 30.0, active_parameter_count_b = 3.0, moe = true, source_url = "https://build.nvidia.com/nvidia/nemotron-3-nano-30b-a3b/modelcard", last_verified = "2026-06-20" }
tier = "mid"
open_weight = true
strengths = ["speed", "cheap", "reasoning", "coding", "tool_use", "long_context"]

[models."nvidia/nemotron-3-nano-omni-30b-a3b-reasoning"]
name = "Nemotron 3 Nano Omni 30B A3B Reasoning (NVIDIA NIM)"
provider = "nvidia"
context_window = 262144
logical_model = "nemotron-3-nano-omni-30b-a3b-reasoning"
equivalence_group = "nemotron-3-nano-omni-30b-a3b-reasoning"
served_variant = "nvidia-nim"
api_dialect = "openai_chat"
capabilities = ["tools", "vision", "video", "audio", "streaming", "thinking"]
architecture = { parameter_count_b = 30.0, active_parameter_count_b = 3.0, moe = true, source_url = "https://build.nvidia.com/nvidia/nemotron-3-nano-omni-30b-a3b-reasoning/modelcard", last_verified = "2026-06-20" }
tier = "mid"
open_weight = true
strengths = ["vision", "reasoning", "tool_use", "summarization", "long_context"]

[models."nvidia/deepseek-v4-pro"]
name = "DeepSeek V4 Pro (NVIDIA NIM)"
provider = "nvidia"
wire_model = "deepseek-ai/deepseek-v4-pro"
context_window = 1048576
logical_model = "deepseek-v4-pro"
equivalence_group = "deepseek-v4-pro"
served_variant = "nvidia-nim"
api_dialect = "openai_chat"
capabilities = ["tools", "streaming", "thinking", "prompt_caching"]
tier = "frontier"
open_weight = true
strengths = ["coding", "agentic", "tool_use", "reasoning", "long_context"]
complementary_with = ["anthropic-claude", "openai-gpt", "google-gemini", "qwen", "kimi"]

[models."nvidia/deepseek-v4-flash"]
name = "DeepSeek V4 Flash (NVIDIA NIM)"
provider = "nvidia"
wire_model = "deepseek-ai/deepseek-v4-flash"
context_window = 1048576
logical_model = "deepseek-v4-flash"
equivalence_group = "deepseek-v4-flash"
served_variant = "nvidia-nim"
api_dialect = "openai_chat"
capabilities = ["tools", "streaming", "thinking", "prompt_caching"]
tier = "mid"
open_weight = true
strengths = ["speed", "coding", "agentic", "tool_use", "reasoning", "long_context"]
complementary_with = ["anthropic-claude", "openai-gpt", "google-gemini", "qwen", "kimi"]

[models."nvidia/minimax-m3"]
name = "MiniMax M3 (NVIDIA NIM)"
provider = "nvidia"
wire_model = "minimaxai/minimax-m3"
context_window = 1048576
logical_model = "minimax-m3"
equivalence_group = "minimax-m3"
served_variant = "nvidia-nim"
api_dialect = "openai_chat"
capabilities = ["tools", "vision", "video", "streaming", "thinking", "prompt_caching"]
pricing = { input_per_mtok = 0.60, output_per_mtok = 2.40, cache_read_per_mtok = 0.12 }
tier = "frontier"
open_weight = false
strengths = ["coding", "agentic", "tool_use", "reasoning", "long_context", "vision"]

[models."nvidia/minimax-m2.7"]
name = "MiniMax M2.7 (NVIDIA NIM)"
provider = "nvidia"
wire_model = "minimaxai/minimax-m2.7"
context_window = 196608
logical_model = "minimax-m2.7"
equivalence_group = "minimax-m2.7"
served_variant = "nvidia-nim"
api_dialect = "openai_chat"
capabilities = ["tools", "streaming", "thinking", "prompt_caching"]
pricing = { input_per_mtok = 0.30, output_per_mtok = 1.20, cache_read_per_mtok = 0.06 }
tier = "frontier"
open_weight = true
strengths = ["speed", "coding", "agentic", "tool_use", "reasoning", "long_context"]

[models."nvidia/kimi-k2.6"]
name = "Kimi K2.6 (NVIDIA NIM)"
provider = "nvidia"
wire_model = "moonshotai/kimi-k2.6"
context_window = 262144
logical_model = "moonshot-kimi-k2.6"
equivalence_group = "moonshot-kimi-k2.6"
served_variant = "nvidia-nim"
api_dialect = "openai_chat"
capabilities = ["tools", "vision", "streaming", "thinking", "prompt_caching"]
pricing = { input_per_mtok = 0.60, output_per_mtok = 2.50, cache_read_per_mtok = 0.15 }
tier = "frontier"
open_weight = true
strengths = ["coding", "agentic", "long_context", "tool_use", "reasoning", "vision"]
complementary_with = ["anthropic-claude", "openai-gpt", "google-gemini", "qwen", "deepseek"]

[models."nvidia/mistral-medium-3.5-128b"]
name = "Mistral Medium 3.5 128B (NVIDIA NIM)"
provider = "nvidia"
wire_model = "mistralai/mistral-medium-3.5-128b"
context_window = 262144
logical_model = "mistral-medium-3.5-128b"
equivalence_group = "mistral-medium-3.5-128b"
served_variant = "nvidia-nim"
api_dialect = "openai_chat"
capabilities = ["tools", "streaming"]
architecture = { parameter_count_b = 128.0, source_url = "https://build.nvidia.com/models", last_verified = "2026-06-20" }
tier = "frontier"
open_weight = true
strengths = ["coding", "agentic", "tool_use"]

[models."nvidia/openai/gpt-oss-120b"]
name = "GPT-OSS 120B (NVIDIA NIM)"
provider = "nvidia"
wire_model = "openai/gpt-oss-120b"
context_window = 131072
logical_model = "openai-gpt-oss-120b"
equivalence_group = "openai-gpt-oss-120b"
served_variant = "nvidia-nim"
api_dialect = "openai_chat"
capabilities = ["tools", "streaming", "thinking"]
architecture = { parameter_count_b = 117.0, active_parameter_count_b = 5.1, moe = true, license = "Apache-2.0", source_url = "https://developers.openai.com/api/docs/models/gpt-oss-120b", last_verified = "2026-06-20" }
tier = "frontier"
open_weight = true
strengths = ["cheap", "tool_use", "reasoning"]

[models."nvidia/openai/gpt-oss-20b"]
name = "GPT-OSS 20B (NVIDIA NIM)"
provider = "nvidia"
wire_model = "openai/gpt-oss-20b"
context_window = 131072
logical_model = "openai-gpt-oss-20b"
equivalence_group = "openai-gpt-oss-20b"
served_variant = "nvidia-nim"
api_dialect = "openai_chat"
capabilities = ["tools", "streaming", "thinking"]
architecture = { parameter_count_b = 21.0, active_parameter_count_b = 3.6, moe = true, license = "Apache-2.0", source_url = "https://developers.openai.com/api/docs/models/gpt-oss-20b", last_verified = "2026-06-20" }
tier = "mid"
open_weight = true
strengths = ["speed", "cheap", "tool_use", "reasoning"]

[models."nvidia/z-ai/glm-5.1"]
name = "GLM 5.1 (NVIDIA NIM)"
provider = "nvidia"
wire_model = "z-ai/glm-5.1"
context_window = 202752
logical_model = "glm-5.1"
equivalence_group = "glm-5.1"
served_variant = "nvidia-nim"
api_dialect = "openai_chat"
capabilities = ["tools", "streaming", "thinking", "prompt_caching"]
pricing = { input_per_mtok = 1.40, output_per_mtok = 4.40, cache_read_per_mtok = 0.26 }
tier = "frontier"
open_weight = true
strengths = ["coding", "agentic", "tool_use", "reasoning", "long_context"]

[models."nvidia/step-3.7-flash"]
name = "Step 3.7 Flash (NVIDIA NIM)"
provider = "nvidia"
wire_model = "stepfun-ai/step-3.7-flash"
context_window = 131072
logical_model = "step-3.7-flash"
equivalence_group = "step-3.7-flash"
served_variant = "nvidia-nim"
api_dialect = "openai_chat"
capabilities = ["tools", "vision", "streaming", "thinking", "prompt_caching"]
tier = "frontier"
open_weight = false
strengths = ["coding", "agentic", "long_context", "tool_use", "reasoning", "vision"]