anomalyco/opencode
1
0
Fork 0
mirror of https://github.com/anomalyco/opencode.git synced 2026-07-23 18:55:37 -04:00
Code Issues 5.9k Packages Projects Releases 19 Wiki Activity

NVIDIA connector is experiencing CRITICAL issues #7768

New Issue
Open
opened 2026-02-16 18:08:10 -05:00 by yindo · 2 comments
yindo commented 2026-02-16 18:08:10 -05:00
Owner
Copy Link
Copy Source

Originally created by @clrblind on GitHub (Jan 27, 2026).

Originally assigned to: @rekram1-node on GitHub.

Description

The connector doesn't cover the entire list of available models, but only a small subset, which is displayed in the open code.
I wrote a small script to query the available models:

import os
import sys
import requests

# Load environment variables if dotenv is available
try:
    from dotenv import load_dotenv
    load_dotenv()
except ImportError:
    pass

NGC_KEY = os.getenv('NGC_KEY')

if not NGC_KEY:
    print("Error: NGC_KEY environment variable is not set.", file=sys.stderr)
    sys.exit(1)

def fetch_models():
    url = 'https://integrate.api.nvidia.com/v1/models'
    headers = {'X-Api-Key': NGC_KEY}
    
    try:
        response = requests.get(url, headers=headers)
        response.raise_for_status()
        data = response.json()
        
        for model in data.get('data', []):
            model_id = model.get('id')
            if model_id:
                print(model_id)
                
    except requests.exceptions.RequestException as e:
        print(f"Error fetching models: {e}", file=sys.stderr)
        sys.exit(1)

if __name__ == "__main__":
    fetch_models()
As a result, I got the following list (spoiler)
  1. 01-ai/yi-large
  2. abacusai/dracarys-llama-3.1-70b-instruct
  3. adept/fuyu-8b
  4. ai21labs/jamba-1.5-large-instruct
  5. ai21labs/jamba-1.5-mini-instruct
  6. aisingapore/sea-lion-7b-instruct
  7. baai/bge-m3
  8. baichuan-inc/baichuan2-13b-chat
  9. bigcode/starcoder2-15b
  10. bigcode/starcoder2-7b
  11. bytedance/seed-oss-36b-instruct
  12. databricks/dbrx-instruct
  13. deepseek-ai/deepseek-coder-6.7b-instruct
  14. deepseek-ai/deepseek-r1-distill-llama-8b
  15. deepseek-ai/deepseek-r1-distill-qwen-14b
  16. deepseek-ai/deepseek-r1-distill-qwen-32b
  17. deepseek-ai/deepseek-r1-distill-qwen-7b
  18. deepseek-ai/deepseek-v3.1
  19. deepseek-ai/deepseek-v3.1-terminus
  20. deepseek-ai/deepseek-v3.2
  21. google/codegemma-1.1-7b
  22. google/codegemma-7b
  23. google/deplot
  24. google/gemma-2-27b-it
  25. google/gemma-2-2b-it
  26. google/gemma-2-9b-it
  27. google/gemma-2b
  28. google/gemma-3-12b-it
  29. google/gemma-3-1b-it
  30. google/gemma-3-27b-it
  31. google/gemma-3-4b-it
  32. google/gemma-3n-e2b-it
  33. google/gemma-3n-e4b-it
  34. google/gemma-7b
  35. google/paligemma
  36. google/recurrentgemma-2b
  37. google/shieldgemma-9b
  38. gotocompany/gemma-2-9b-cpt-sahabatai-instruct
  39. ibm/granite-3.0-3b-a800m-instruct
  40. ibm/granite-3.0-8b-instruct
  41. ibm/granite-3.3-8b-instruct
  42. ibm/granite-34b-code-instruct
  43. ibm/granite-8b-code-instruct
  44. ibm/granite-guardian-3.0-8b
  45. igenius/colosseum_355b_instruct_16k
  46. igenius/italia_10b_instruct_16k
  47. institute-of-science-tokyo/llama-3.1-swallow-70b-instruct-v0.1
  48. institute-of-science-tokyo/llama-3.1-swallow-8b-instruct-v0.1
  49. marin/marin-8b-instruct
  50. mediatek/breeze-7b-instruct
  51. meta/codellama-70b
  52. meta/llama-3.1-405b-instruct
  53. meta/llama-3.1-70b-instruct
  54. meta/llama-3.1-8b-instruct
  55. meta/llama-3.2-11b-vision-instruct
  56. meta/llama-3.2-1b-instruct
  57. meta/llama-3.2-3b-instruct
  58. meta/llama-3.2-90b-vision-instruct
  59. meta/llama-3.3-70b-instruct
  60. meta/llama-4-maverick-17b-128e-instruct
  61. meta/llama-4-scout-17b-16e-instruct
  62. meta/llama-guard-4-12b
  63. meta/llama2-70b
  64. meta/llama3-70b-instruct
  65. meta/llama3-8b-instruct
  66. microsoft/kosmos-2
  67. microsoft/phi-3-medium-128k-instruct
  68. microsoft/phi-3-medium-4k-instruct
  69. microsoft/phi-3-mini-128k-instruct
  70. microsoft/phi-3-mini-4k-instruct
  71. microsoft/phi-3-small-128k-instruct
  72. microsoft/phi-3-small-8k-instruct
  73. microsoft/phi-3-vision-128k-instruct
  74. microsoft/phi-3.5-mini-instruct
  75. microsoft/phi-3.5-moe-instruct
  76. microsoft/phi-3.5-vision-instruct
  77. microsoft/phi-4-mini-flash-reasoning
  78. microsoft/phi-4-mini-instruct
  79. microsoft/phi-4-multimodal-instruct
  80. minimaxai/minimax-m2
  81. minimaxai/minimax-m2.1
  82. mistralai/codestral-22b-instruct-v0.1
  83. mistralai/devstral-2-123b-instruct-2512
  84. mistralai/magistral-small-2506
  85. mistralai/mamba-codestral-7b-v0.1
  86. mistralai/mathstral-7b-v0.1
  87. mistralai/ministral-14b-instruct-2512
  88. mistralai/mistral-7b-instruct-v0.2
  89. mistralai/mistral-7b-instruct-v0.3
  90. mistralai/mistral-large
  91. mistralai/mistral-large-2-instruct
  92. mistralai/mistral-large-3-675b-instruct-2512
  93. mistralai/mistral-medium-3-instruct
  94. mistralai/mistral-nemotron
  95. mistralai/mistral-small-24b-instruct
  96. mistralai/mistral-small-3.1-24b-instruct-2503
  97. mistralai/mixtral-8x22b-instruct-v0.1
  98. mistralai/mixtral-8x22b-v0.1
  99. mistralai/mixtral-8x7b-instruct-v0.1
  100. moonshotai/kimi-k2-instruct
  101. moonshotai/kimi-k2-instruct-0905
  102. moonshotai/kimi-k2-thinking
  103. moonshotai/kimi-k2.5
  104. nv-mistralai/mistral-nemo-12b-instruct
  105. nvidia/cosmos-reason2-8b
  106. nvidia/embed-qa-4
  107. nvidia/llama-3.1-nemoguard-8b-content-safety
  108. nvidia/llama-3.1-nemoguard-8b-topic-control
  109. nvidia/llama-3.1-nemotron-51b-instruct
  110. nvidia/llama-3.1-nemotron-70b-instruct
  111. nvidia/llama-3.1-nemotron-70b-reward
  112. nvidia/llama-3.1-nemotron-nano-4b-v1.1
  113. nvidia/llama-3.1-nemotron-nano-8b-v1
  114. nvidia/llama-3.1-nemotron-nano-vl-8b-v1
  115. nvidia/llama-3.1-nemotron-safety-guard-8b-v3
  116. nvidia/llama-3.1-nemotron-ultra-253b-v1
  117. nvidia/llama-3.2-nemoretriever-1b-vlm-embed-v1
  118. nvidia/llama-3.2-nemoretriever-300m-embed-v1
  119. nvidia/llama-3.2-nemoretriever-300m-embed-v2
  120. nvidia/llama-3.2-nv-embedqa-1b-v1
  121. nvidia/llama-3.2-nv-embedqa-1b-v2
  122. nvidia/llama-3.3-nemotron-super-49b-v1
  123. nvidia/llama-3.3-nemotron-super-49b-v1.5
  124. nvidia/llama3-chatqa-1.5-70b
  125. nvidia/llama3-chatqa-1.5-8b
  126. nvidia/mistral-nemo-minitron-8b-8k-instruct
  127. nvidia/mistral-nemo-minitron-8b-base
  128. nvidia/nemoretriever-parse
  129. nvidia/nemotron-3-nano-30b-a3b
  130. nvidia/nemotron-4-340b-instruct
  131. nvidia/nemotron-4-340b-reward
  132. nvidia/nemotron-4-mini-hindi-4b-instruct
  133. nvidia/nemotron-content-safety-reasoning-4b
  134. nvidia/nemotron-mini-4b-instruct
  135. nvidia/nemotron-nano-12b-v2-vl
  136. nvidia/nemotron-nano-3-30b-a3b
  137. nvidia/nemotron-parse
  138. nvidia/neva-22b
  139. nvidia/nv-embed-v1
  140. nvidia/nv-embedcode-7b-v1
  141. nvidia/nv-embedqa-e5-v5
  142. nvidia/nv-embedqa-mistral-7b-v2
  143. nvidia/nvclip
  144. nvidia/nvidia-nemotron-nano-9b-v2
  145. nvidia/riva-translate-4b-instruct
  146. nvidia/riva-translate-4b-instruct-v1.1
  147. nvidia/streampetr
  148. nvidia/usdcode-llama-3.1-70b-instruct
  149. nvidia/vila
  150. openai/gpt-oss-120b
  151. openai/gpt-oss-120b
  152. openai/gpt-oss-20b
  153. openai/gpt-oss-20b
  154. opengpt-x/teuken-7b-instruct-commercial-v0.4
  155. qwen/qwen2-7b-instruct
  156. qwen/qwen2.5-7b-instruct
  157. qwen/qwen2.5-coder-32b-instruct
  158. qwen/qwen2.5-coder-7b-instruct
  159. qwen/qwen3-235b-a22b
  160. qwen/qwen3-coder-480b-a35b-instruct
  161. qwen/qwen3-next-80b-a3b-instruct
  162. qwen/qwen3-next-80b-a3b-thinking
  163. qwen/qwq-32b
  164. rakuten/rakutenai-7b-chat
  165. rakuten/rakutenai-7b-instruct
  166. sarvamai/sarvam-m
  167. snowflake/arctic-embed-l
  168. speakleash/bielik-11b-v2.3-instruct
  169. speakleash/bielik-11b-v2.6-instruct
  170. stockmark/stockmark-2-100b-instruct
  171. thudm/chatglm3-6b
  172. tiiuae/falcon3-7b-instruct
  173. tokyotech-llm/llama-3-swallow-70b-instruct-v0.1
  174. upstage/solar-10.7b-instruct
  175. utter-project/eurollm-9b-instruct
  176. writer/palmyra-creative-122b
  177. writer/palmyra-fin-70b-32k
  178. writer/palmyra-med-70b
  179. writer/palmyra-med-70b-32k
  180. yentinglin/llama-3-taiwan-70b-instruct
  181. z-ai/glm4.7
  182. zyphra/zamba2-7b-instruct

Yes, not all of them support text input, for example nvidia/nemotron-parse, but nevertheless I surveyed each of the models and based on the survey results

received successful responses from 116 models. (spoiler)

abacusai/dracarys-llama-3.1-70b-instruct: 1.24s
ai21labs/jamba-1.5-mini-instruct: 1.24s
deepseek-ai/deepseek-r1-distill-qwen-7b: 1.42s
deepseek-ai/deepseek-r1-distill-qwen-14b: 1.70s
google/gemma-3-1b-it: 0.85s
bytedance/seed-oss-36b-instruct: 2.11s
google/gemma-2-2b-it: 1.42s
google/paligemma: Error: 500 (0.97s)
google/shieldgemma-9b: 0.85s
google/gemma-2-27b-it: 1.85s
google/gemma-2-9b-it: 1.89s
google/gemma-3-27b-it: 1.64s
baichuan-inc/baichuan2-13b-chat: 2.99s
gotocompany/gemma-2-9b-cpt-sahabatai-instruct: 1.17s
google/gemma-7b: 2.04s
ibm/granite-guardian-3.0-8b: 0.99s
igenius/italia_10b_instruct_16k: 0.92s
institute-of-science-tokyo/llama-3.1-swallow-70b-instruct-v0.1: 1.07s
institute-of-science-tokyo/llama-3.1-swallow-8b-instruct-v0.1: 1.05s
google/gemma-3n-e2b-it: 2.89s
deepseek-ai/deepseek-v3.1: 4.17s
marin/marin-8b-instruct: 1.31s
mediatek/breeze-7b-instruct: 1.39s
meta/llama-3.2-1b-instruct: 0.77s
meta/llama-3.2-3b-instruct: 0.66s
meta/llama-3.1-8b-instruct: 1.06s
ibm/granite-3.3-8b-instruct: 2.18s
google/gemma-3n-e4b-it: 3.48s
meta/llama-3.1-70b-instruct: 1.35s
meta/llama-3.1-405b-instruct: 1.79s
igenius/colosseum_355b_instruct_16k: 2.45s
meta/llama-4-scout-17b-16e-instruct: 0.67s
meta/llama-guard-4-12b: 0.67s
meta/llama-4-maverick-17b-128e-instruct: 1.13s
meta/llama-3.2-11b-vision-instruct: 1.75s
meta/llama-3.2-90b-vision-instruct: 1.52s
meta/llama-3.3-70b-instruct: 1.47s
meta/llama3-8b-instruct: 1.20s
google/gemma-3-4b-it: 4.75s
meta/llama3-70b-instruct: 1.79s
microsoft/phi-3-small-128k-instruct: 1.33s
microsoft/phi-3-medium-4k-instruct: 1.61s
microsoft/phi-3-mini-4k-instruct: 1.51s
microsoft/phi-3.5-vision-instruct: 1.03s
microsoft/phi-3-small-8k-instruct: 1.59s
microsoft/phi-4-mini-flash-reasoning: 1.28s
microsoft/phi-3-mini-128k-instruct: 2.04s
microsoft/phi-3.5-mini-instruct: 1.93s
minimaxai/minimax-m2: 1.58s
google/gemma-3-12b-it: 6.65s
microsoft/phi-3-medium-128k-instruct: 2.96s
mistralai/ministral-14b-instruct-2512: 1.33s
minimaxai/minimax-m2.1: 2.06s
mistralai/mathstral-7b-v0.1: 1.54s
mistralai/mistral-7b-instruct-v0.2: 1.50s
mistralai/mistral-7b-instruct-v0.3: 1.72s
mistralai/mamba-codestral-7b-v0.1: 2.03s
mistralai/devstral-2-123b-instruct-2512: 2.30s
mistralai/magistral-small-2506: 2.33s
moonshotai/kimi-k2-instruct: 1.25s
mistralai/mistral-nemotron: 1.96s
mistralai/mixtral-8x7b-instruct-v0.1: 1.51s
moonshotai/kimi-k2-instruct-0905: 1.38s
mistralai/mistral-medium-3-instruct: 2.22s
mistralai/mixtral-8x22b-instruct-v0.1: 1.91s
moonshotai/kimi-k2-thinking: 1.54s
nvidia/llama-3.1-nemoguard-8b-topic-control: 0.64s
nvidia/llama-3.1-nemoguard-8b-content-safety: 0.81s
mistralai/mistral-small-24b-instruct: 2.54s
mistralai/mistral-small-3.1-24b-instruct-2503: 2.58s
nvidia/llama-3.1-nemotron-safety-guard-8b-v3: 0.70s
nvidia/llama-3.1-nemotron-nano-4b-v1.1: 1.25s
nvidia/llama-3.1-nemotron-nano-vl-8b-v1: 1.34s
nvidia/llama-3.1-nemotron-nano-8b-v1: 1.43s
nvidia/llama3-chatqa-1.5-8b: 0.89s
nvidia/llama-3.1-nemotron-70b-reward: 1.98s
nvidia/nemoretriever-parse: Error: 400 (0.80s)
nvidia/llama-3.1-nemotron-ultra-253b-v1: 1.87s
nvidia/nemotron-3-nano-30b-a3b: 1.12s
nvidia/nemotron-content-safety-reasoning-4b: 0.87s
nvidia/nemotron-4-mini-hindi-4b-instruct: 1.12s
nvidia/nemotron-mini-4b-instruct: 1.10s
nvidia/nemotron-parse: Error: 400 (0.87s)
nvidia/llama-3.3-nemotron-super-49b-v1.5: 2.20s
nvidia/llama-3.3-nemotron-super-49b-v1: 2.29s
nvidia/nemotron-nano-12b-v2-vl: 1.50s
openai/gpt-oss-20b: 0.80s
openai/gpt-oss-120b: 1.03s
openai/gpt-oss-120b: 1.22s
nvidia/riva-translate-4b-instruct-v1.1: 1.30s
openai/gpt-oss-20b: 1.22s
qwen/qwen2-7b-instruct: 1.20s
qwen/qwen2.5-7b-instruct: 1.23s
qwen/qwen2.5-coder-7b-instruct: 0.97s
opengpt-x/teuken-7b-instruct-commercial-v0.4: 1.80s
qwen/qwen3-235b-a22b: 1.84s
qwen/qwen2.5-coder-32b-instruct: 2.22s
rakuten/rakutenai-7b-instruct: 1.67s
rakuten/rakutenai-7b-chat: 2.02s
nvidia/usdcode-llama-3.1-70b-instruct: 4.11s
sarvamai/sarvam-m: 2.27s
speakleash/bielik-11b-v2.3-instruct: 1.62s
moonshotai/kimi-k2.5: 8.52s
deepseek-ai/deepseek-v3.1-terminus: 17.10s
speakleash/bielik-11b-v2.6-instruct: 2.17s
qwen/qwen3-coder-480b-a35b-instruct: 3.95s
thudm/chatglm3-6b: 1.57s
tiiuae/falcon3-7b-instruct: 1.33s
stockmark/stockmark-2-100b-instruct: 2.66s
utter-project/eurollm-9b-instruct: 1.44s
upstage/solar-10.7b-instruct: 2.36s
tokyotech-llm/llama-3-swallow-70b-instruct-v0.1: 3.24s
yentinglin/llama-3-taiwan-70b-instruct: 2.37s
nvidia/nvidia-nemotron-nano-9b-v2: 9.11s
z-ai/glm4.7: 26.84s
microsoft/phi-4-multimodal-instruct: 58.71s
qwen/qwen3-next-80b-a3b-thinking: 183.53s
qwen/qwen3-next-80b-a3b-instruct: 266.68s
mistralai/mistral-large-3-675b-instruct-2512: 293.33s
deepseek-ai/deepseek-r1-distill-llama-8b: Error: Timeout (300.00s)
deepseek-ai/deepseek-v3.2: Error: Timeout (300.00s)
deepseek-ai/deepseek-r1-distill-qwen-32b: Error: Timeout (300.00s)
microsoft/phi-4-mini-instruct: Error: Timeout (300.00s)
qwen/qwq-32b: Error: Timeout (300.00s)

Yes, large models like the Qwen3 and DeepSeek R1 didn't respond and were timed out, but the very high-quality GLM model did respond. If you select the GLM model in the open code, it simply doesn't respond, although I personally verified that it works fine using the API.

I urge you to look into this issue, as this connector is a gem in vibe coding.

I also found a problem with some llama models in opencode, instead of calling tools they send a <tool_call> message and this is not handled in any way:

Check file VERSION.md
11:33 PM
Thinking: Okay, the user wants me to check the contents of the VERSION.md file. Let me start by recalling the available tools. The 'read' tool is used to read files from the local filesystem. The parameters require an absolute file path. Since the user mentioned VERSION.md, I need to confirm the absolute path.
Looking at the environment details, the working directory is /home/albert/Desktop/mytest. So the absolute path would be /home/albert/Desktop/mytest/VERSION.md. I should use the read tool with this file path. There's no mention of needing offset or limit, so I'll omit those parameters. Let me make sure that the file exists. Wait, the user didn't specify, but the read tool can handle non-existent files by returning an error. But since the user is asking to check it, it's likely that the file exists. I'll proceed to call the read tool with the specified path.
<tool_call>
{"name": "read", "arguments": {"filePath": "/home/user/mytest/VERSION.md"}}
</tool_call>

Due to this critical issue, the model becomes completely unsuitable for any interaction, since it can only effectively respond in text.

Plugins

none

OpenCode version

1.1.36

Steps to reproduce

No response

Screenshot and/or share link

No response

Operating System

Linux Mint 22.1 x86_64

Terminal

gnome-terminal / bash 5.2.21

Originally created by @clrblind on GitHub (Jan 27, 2026). Originally assigned to: @rekram1-node on GitHub. ### Description The connector doesn't cover the entire list of available models, but only a small subset, which is displayed in the open code. I wrote a small script to query the available models: ```python import os import sys import requests # Load environment variables if dotenv is available try: from dotenv import load_dotenv load_dotenv() except ImportError: pass NGC_KEY = os.getenv('NGC_KEY') if not NGC_KEY: print("Error: NGC_KEY environment variable is not set.", file=sys.stderr) sys.exit(1) def fetch_models(): url = 'https://integrate.api.nvidia.com/v1/models' headers = {'X-Api-Key': NGC_KEY} try: response = requests.get(url, headers=headers) response.raise_for_status() data = response.json() for model in data.get('data', []): model_id = model.get('id') if model_id: print(model_id) except requests.exceptions.RequestException as e: print(f"Error fetching models: {e}", file=sys.stderr) sys.exit(1) if __name__ == "__main__": fetch_models() ``` <details> <summary>As a result, I got the following list (spoiler)</summary> 1. 01-ai/yi-large 2. abacusai/dracarys-llama-3.1-70b-instruct 3. adept/fuyu-8b 4. ai21labs/jamba-1.5-large-instruct 5. ai21labs/jamba-1.5-mini-instruct 6. aisingapore/sea-lion-7b-instruct 7. baai/bge-m3 8. baichuan-inc/baichuan2-13b-chat 9. bigcode/starcoder2-15b 10. bigcode/starcoder2-7b 11. bytedance/seed-oss-36b-instruct 12. databricks/dbrx-instruct 13. deepseek-ai/deepseek-coder-6.7b-instruct 14. deepseek-ai/deepseek-r1-distill-llama-8b 15. deepseek-ai/deepseek-r1-distill-qwen-14b 16. deepseek-ai/deepseek-r1-distill-qwen-32b 17. deepseek-ai/deepseek-r1-distill-qwen-7b 18. deepseek-ai/deepseek-v3.1 19. deepseek-ai/deepseek-v3.1-terminus 20. deepseek-ai/deepseek-v3.2 21. google/codegemma-1.1-7b 22. google/codegemma-7b 23. google/deplot 24. google/gemma-2-27b-it 25. google/gemma-2-2b-it 26. google/gemma-2-9b-it 27. google/gemma-2b 28. google/gemma-3-12b-it 29. google/gemma-3-1b-it 30. google/gemma-3-27b-it 31. google/gemma-3-4b-it 32. google/gemma-3n-e2b-it 33. google/gemma-3n-e4b-it 34. google/gemma-7b 35. google/paligemma 36. google/recurrentgemma-2b 37. google/shieldgemma-9b 38. gotocompany/gemma-2-9b-cpt-sahabatai-instruct 39. ibm/granite-3.0-3b-a800m-instruct 40. ibm/granite-3.0-8b-instruct 41. ibm/granite-3.3-8b-instruct 42. ibm/granite-34b-code-instruct 43. ibm/granite-8b-code-instruct 44. ibm/granite-guardian-3.0-8b 45. igenius/colosseum_355b_instruct_16k 46. igenius/italia_10b_instruct_16k 47. institute-of-science-tokyo/llama-3.1-swallow-70b-instruct-v0.1 48. institute-of-science-tokyo/llama-3.1-swallow-8b-instruct-v0.1 49. marin/marin-8b-instruct 50. mediatek/breeze-7b-instruct 51. meta/codellama-70b 52. meta/llama-3.1-405b-instruct 53. meta/llama-3.1-70b-instruct 54. meta/llama-3.1-8b-instruct 55. meta/llama-3.2-11b-vision-instruct 56. meta/llama-3.2-1b-instruct 57. meta/llama-3.2-3b-instruct 58. meta/llama-3.2-90b-vision-instruct 59. meta/llama-3.3-70b-instruct 60. meta/llama-4-maverick-17b-128e-instruct 61. meta/llama-4-scout-17b-16e-instruct 62. meta/llama-guard-4-12b 63. meta/llama2-70b 64. meta/llama3-70b-instruct 65. meta/llama3-8b-instruct 66. microsoft/kosmos-2 67. microsoft/phi-3-medium-128k-instruct 68. microsoft/phi-3-medium-4k-instruct 69. microsoft/phi-3-mini-128k-instruct 70. microsoft/phi-3-mini-4k-instruct 71. microsoft/phi-3-small-128k-instruct 72. microsoft/phi-3-small-8k-instruct 73. microsoft/phi-3-vision-128k-instruct 74. microsoft/phi-3.5-mini-instruct 75. microsoft/phi-3.5-moe-instruct 76. microsoft/phi-3.5-vision-instruct 77. microsoft/phi-4-mini-flash-reasoning 78. microsoft/phi-4-mini-instruct 79. microsoft/phi-4-multimodal-instruct 80. minimaxai/minimax-m2 81. minimaxai/minimax-m2.1 82. mistralai/codestral-22b-instruct-v0.1 83. mistralai/devstral-2-123b-instruct-2512 84. mistralai/magistral-small-2506 85. mistralai/mamba-codestral-7b-v0.1 86. mistralai/mathstral-7b-v0.1 87. mistralai/ministral-14b-instruct-2512 88. mistralai/mistral-7b-instruct-v0.2 89. mistralai/mistral-7b-instruct-v0.3 90. mistralai/mistral-large 91. mistralai/mistral-large-2-instruct 92. mistralai/mistral-large-3-675b-instruct-2512 93. mistralai/mistral-medium-3-instruct 94. mistralai/mistral-nemotron 95. mistralai/mistral-small-24b-instruct 96. mistralai/mistral-small-3.1-24b-instruct-2503 97. mistralai/mixtral-8x22b-instruct-v0.1 98. mistralai/mixtral-8x22b-v0.1 99. mistralai/mixtral-8x7b-instruct-v0.1 100. moonshotai/kimi-k2-instruct 101. moonshotai/kimi-k2-instruct-0905 102. moonshotai/kimi-k2-thinking 103. moonshotai/kimi-k2.5 104. nv-mistralai/mistral-nemo-12b-instruct 105. nvidia/cosmos-reason2-8b 106. nvidia/embed-qa-4 107. nvidia/llama-3.1-nemoguard-8b-content-safety 108. nvidia/llama-3.1-nemoguard-8b-topic-control 109. nvidia/llama-3.1-nemotron-51b-instruct 110. nvidia/llama-3.1-nemotron-70b-instruct 111. nvidia/llama-3.1-nemotron-70b-reward 112. nvidia/llama-3.1-nemotron-nano-4b-v1.1 113. nvidia/llama-3.1-nemotron-nano-8b-v1 114. nvidia/llama-3.1-nemotron-nano-vl-8b-v1 115. nvidia/llama-3.1-nemotron-safety-guard-8b-v3 116. nvidia/llama-3.1-nemotron-ultra-253b-v1 117. nvidia/llama-3.2-nemoretriever-1b-vlm-embed-v1 118. nvidia/llama-3.2-nemoretriever-300m-embed-v1 119. nvidia/llama-3.2-nemoretriever-300m-embed-v2 120. nvidia/llama-3.2-nv-embedqa-1b-v1 121. nvidia/llama-3.2-nv-embedqa-1b-v2 122. nvidia/llama-3.3-nemotron-super-49b-v1 123. nvidia/llama-3.3-nemotron-super-49b-v1.5 124. nvidia/llama3-chatqa-1.5-70b 125. nvidia/llama3-chatqa-1.5-8b 126. nvidia/mistral-nemo-minitron-8b-8k-instruct 127. nvidia/mistral-nemo-minitron-8b-base 128. nvidia/nemoretriever-parse 129. nvidia/nemotron-3-nano-30b-a3b 130. nvidia/nemotron-4-340b-instruct 131. nvidia/nemotron-4-340b-reward 132. nvidia/nemotron-4-mini-hindi-4b-instruct 133. nvidia/nemotron-content-safety-reasoning-4b 134. nvidia/nemotron-mini-4b-instruct 135. nvidia/nemotron-nano-12b-v2-vl 136. nvidia/nemotron-nano-3-30b-a3b 137. nvidia/nemotron-parse 138. nvidia/neva-22b 139. nvidia/nv-embed-v1 140. nvidia/nv-embedcode-7b-v1 141. nvidia/nv-embedqa-e5-v5 142. nvidia/nv-embedqa-mistral-7b-v2 143. nvidia/nvclip 144. nvidia/nvidia-nemotron-nano-9b-v2 145. nvidia/riva-translate-4b-instruct 146. nvidia/riva-translate-4b-instruct-v1.1 147. nvidia/streampetr 148. nvidia/usdcode-llama-3.1-70b-instruct 149. nvidia/vila 150. openai/gpt-oss-120b 151. openai/gpt-oss-120b 152. openai/gpt-oss-20b 153. openai/gpt-oss-20b 154. opengpt-x/teuken-7b-instruct-commercial-v0.4 155. qwen/qwen2-7b-instruct 156. qwen/qwen2.5-7b-instruct 157. qwen/qwen2.5-coder-32b-instruct 158. qwen/qwen2.5-coder-7b-instruct 159. qwen/qwen3-235b-a22b 160. qwen/qwen3-coder-480b-a35b-instruct 161. qwen/qwen3-next-80b-a3b-instruct 162. qwen/qwen3-next-80b-a3b-thinking 163. qwen/qwq-32b 164. rakuten/rakutenai-7b-chat 165. rakuten/rakutenai-7b-instruct 166. sarvamai/sarvam-m 167. snowflake/arctic-embed-l 168. speakleash/bielik-11b-v2.3-instruct 169. speakleash/bielik-11b-v2.6-instruct 170. stockmark/stockmark-2-100b-instruct 171. thudm/chatglm3-6b 172. tiiuae/falcon3-7b-instruct 173. tokyotech-llm/llama-3-swallow-70b-instruct-v0.1 174. upstage/solar-10.7b-instruct 175. utter-project/eurollm-9b-instruct 176. writer/palmyra-creative-122b 177. writer/palmyra-fin-70b-32k 178. writer/palmyra-med-70b 179. writer/palmyra-med-70b-32k 180. yentinglin/llama-3-taiwan-70b-instruct 181. z-ai/glm4.7 182. zyphra/zamba2-7b-instruct </details> Yes, not all of them support text input, for example nvidia/nemotron-parse, but nevertheless I surveyed each of the models and based on the survey results <details> <summary>received successful responses from 116 models. (spoiler)</summary> abacusai/dracarys-llama-3.1-70b-instruct: 1.24s ai21labs/jamba-1.5-mini-instruct: 1.24s deepseek-ai/deepseek-r1-distill-qwen-7b: 1.42s deepseek-ai/deepseek-r1-distill-qwen-14b: 1.70s google/gemma-3-1b-it: 0.85s bytedance/seed-oss-36b-instruct: 2.11s google/gemma-2-2b-it: 1.42s google/paligemma: Error: 500 (0.97s) google/shieldgemma-9b: 0.85s google/gemma-2-27b-it: 1.85s google/gemma-2-9b-it: 1.89s google/gemma-3-27b-it: 1.64s baichuan-inc/baichuan2-13b-chat: 2.99s gotocompany/gemma-2-9b-cpt-sahabatai-instruct: 1.17s google/gemma-7b: 2.04s ibm/granite-guardian-3.0-8b: 0.99s igenius/italia_10b_instruct_16k: 0.92s institute-of-science-tokyo/llama-3.1-swallow-70b-instruct-v0.1: 1.07s institute-of-science-tokyo/llama-3.1-swallow-8b-instruct-v0.1: 1.05s google/gemma-3n-e2b-it: 2.89s deepseek-ai/deepseek-v3.1: 4.17s marin/marin-8b-instruct: 1.31s mediatek/breeze-7b-instruct: 1.39s meta/llama-3.2-1b-instruct: 0.77s meta/llama-3.2-3b-instruct: 0.66s meta/llama-3.1-8b-instruct: 1.06s ibm/granite-3.3-8b-instruct: 2.18s google/gemma-3n-e4b-it: 3.48s meta/llama-3.1-70b-instruct: 1.35s meta/llama-3.1-405b-instruct: 1.79s igenius/colosseum_355b_instruct_16k: 2.45s meta/llama-4-scout-17b-16e-instruct: 0.67s meta/llama-guard-4-12b: 0.67s meta/llama-4-maverick-17b-128e-instruct: 1.13s meta/llama-3.2-11b-vision-instruct: 1.75s meta/llama-3.2-90b-vision-instruct: 1.52s meta/llama-3.3-70b-instruct: 1.47s meta/llama3-8b-instruct: 1.20s google/gemma-3-4b-it: 4.75s meta/llama3-70b-instruct: 1.79s microsoft/phi-3-small-128k-instruct: 1.33s microsoft/phi-3-medium-4k-instruct: 1.61s microsoft/phi-3-mini-4k-instruct: 1.51s microsoft/phi-3.5-vision-instruct: 1.03s microsoft/phi-3-small-8k-instruct: 1.59s microsoft/phi-4-mini-flash-reasoning: 1.28s microsoft/phi-3-mini-128k-instruct: 2.04s microsoft/phi-3.5-mini-instruct: 1.93s minimaxai/minimax-m2: 1.58s google/gemma-3-12b-it: 6.65s microsoft/phi-3-medium-128k-instruct: 2.96s mistralai/ministral-14b-instruct-2512: 1.33s minimaxai/minimax-m2.1: 2.06s mistralai/mathstral-7b-v0.1: 1.54s mistralai/mistral-7b-instruct-v0.2: 1.50s mistralai/mistral-7b-instruct-v0.3: 1.72s mistralai/mamba-codestral-7b-v0.1: 2.03s mistralai/devstral-2-123b-instruct-2512: 2.30s mistralai/magistral-small-2506: 2.33s moonshotai/kimi-k2-instruct: 1.25s mistralai/mistral-nemotron: 1.96s mistralai/mixtral-8x7b-instruct-v0.1: 1.51s moonshotai/kimi-k2-instruct-0905: 1.38s mistralai/mistral-medium-3-instruct: 2.22s mistralai/mixtral-8x22b-instruct-v0.1: 1.91s moonshotai/kimi-k2-thinking: 1.54s nvidia/llama-3.1-nemoguard-8b-topic-control: 0.64s nvidia/llama-3.1-nemoguard-8b-content-safety: 0.81s mistralai/mistral-small-24b-instruct: 2.54s mistralai/mistral-small-3.1-24b-instruct-2503: 2.58s nvidia/llama-3.1-nemotron-safety-guard-8b-v3: 0.70s nvidia/llama-3.1-nemotron-nano-4b-v1.1: 1.25s nvidia/llama-3.1-nemotron-nano-vl-8b-v1: 1.34s nvidia/llama-3.1-nemotron-nano-8b-v1: 1.43s nvidia/llama3-chatqa-1.5-8b: 0.89s nvidia/llama-3.1-nemotron-70b-reward: 1.98s nvidia/nemoretriever-parse: Error: 400 (0.80s) nvidia/llama-3.1-nemotron-ultra-253b-v1: 1.87s nvidia/nemotron-3-nano-30b-a3b: 1.12s nvidia/nemotron-content-safety-reasoning-4b: 0.87s nvidia/nemotron-4-mini-hindi-4b-instruct: 1.12s nvidia/nemotron-mini-4b-instruct: 1.10s nvidia/nemotron-parse: Error: 400 (0.87s) nvidia/llama-3.3-nemotron-super-49b-v1.5: 2.20s nvidia/llama-3.3-nemotron-super-49b-v1: 2.29s nvidia/nemotron-nano-12b-v2-vl: 1.50s openai/gpt-oss-20b: 0.80s openai/gpt-oss-120b: 1.03s openai/gpt-oss-120b: 1.22s nvidia/riva-translate-4b-instruct-v1.1: 1.30s openai/gpt-oss-20b: 1.22s qwen/qwen2-7b-instruct: 1.20s qwen/qwen2.5-7b-instruct: 1.23s qwen/qwen2.5-coder-7b-instruct: 0.97s opengpt-x/teuken-7b-instruct-commercial-v0.4: 1.80s qwen/qwen3-235b-a22b: 1.84s qwen/qwen2.5-coder-32b-instruct: 2.22s rakuten/rakutenai-7b-instruct: 1.67s rakuten/rakutenai-7b-chat: 2.02s nvidia/usdcode-llama-3.1-70b-instruct: 4.11s sarvamai/sarvam-m: 2.27s speakleash/bielik-11b-v2.3-instruct: 1.62s moonshotai/kimi-k2.5: 8.52s deepseek-ai/deepseek-v3.1-terminus: 17.10s speakleash/bielik-11b-v2.6-instruct: 2.17s qwen/qwen3-coder-480b-a35b-instruct: 3.95s thudm/chatglm3-6b: 1.57s tiiuae/falcon3-7b-instruct: 1.33s stockmark/stockmark-2-100b-instruct: 2.66s utter-project/eurollm-9b-instruct: 1.44s upstage/solar-10.7b-instruct: 2.36s tokyotech-llm/llama-3-swallow-70b-instruct-v0.1: 3.24s yentinglin/llama-3-taiwan-70b-instruct: 2.37s nvidia/nvidia-nemotron-nano-9b-v2: 9.11s z-ai/glm4.7: 26.84s microsoft/phi-4-multimodal-instruct: 58.71s qwen/qwen3-next-80b-a3b-thinking: 183.53s qwen/qwen3-next-80b-a3b-instruct: 266.68s mistralai/mistral-large-3-675b-instruct-2512: 293.33s deepseek-ai/deepseek-r1-distill-llama-8b: Error: Timeout (300.00s) deepseek-ai/deepseek-v3.2: Error: Timeout (300.00s) deepseek-ai/deepseek-r1-distill-qwen-32b: Error: Timeout (300.00s) microsoft/phi-4-mini-instruct: Error: Timeout (300.00s) qwen/qwq-32b: Error: Timeout (300.00s) </details> Yes, large models like the Qwen3 and DeepSeek R1 didn't respond and were timed out, but the very high-quality GLM model did respond. If you select the GLM model in the open code, it simply doesn't respond, although I personally verified that it works fine using the API. I urge you to look into this issue, as this connector is a gem in vibe coding. I also found a problem with some llama models in opencode, instead of calling tools they send a <tool_call> message and this is not handled in any way: > Check file VERSION.md > 11:33 PM > Thinking: Okay, the user wants me to check the contents of the VERSION.md file. Let me start by recalling the available tools. The 'read' tool is used to read files from the local filesystem. The parameters require an absolute file path. Since the user mentioned VERSION.md, I need to confirm the absolute path. > Looking at the environment details, the working directory is /home/albert/Desktop/mytest. So the absolute path would be /home/albert/Desktop/mytest/VERSION.md. I should use the read tool with this file path. There's no mention of needing offset or limit, so I'll omit those parameters. Let me make sure that the file exists. Wait, the user didn't specify, but the read tool can handle non-existent files by returning an error. But since the user is asking to check it, it's likely that the file exists. I'll proceed to call the read tool with the specified path. > <tool_call> > {"name": "read", "arguments": {"filePath": "/home/user/mytest/VERSION.md"}} > </tool_call> Due to this critical issue, the model becomes completely unsuitable for any interaction, since it can only effectively respond in text. ### Plugins none ### OpenCode version 1.1.36 ### Steps to reproduce _No response_ ### Screenshot and/or share link _No response_ ### Operating System Linux Mint 22.1 x86_64 ### Terminal gnome-terminal / bash 5.2.21
yindo added the bugzen labels 2026-02-16 18:08:10 -05:00
yindo commented 2026-02-16 18:08:11 -05:00
Author
Owner
Copy Link
Copy Source

@github-actions[bot] commented on GitHub (Jan 27, 2026):

This issue might be a duplicate of existing issues. Please check:

  • #10855: Qwen3 Coder on OpenCode Zen: Tool call metadata being printed to chat instead of actually being called (relates to the <tool_call> message handling issue you mentioned)
  • #10673: Tool call interruption issues with specific models leading to conversation stops
  • #10687: Models not showing up in UI despite being available in configuration (relates to the incomplete model list issue)

Feel free to ignore if none of these address your specific case.

@github-actions[bot] commented on GitHub (Jan 27, 2026): This issue might be a duplicate of existing issues. Please check: - #10855: Qwen3 Coder on OpenCode Zen: Tool call metadata being printed to chat instead of actually being called (relates to the `<tool_call>` message handling issue you mentioned) - #10673: Tool call interruption issues with specific models leading to conversation stops - #10687: Models not showing up in UI despite being available in configuration (relates to the incomplete model list issue) Feel free to ignore if none of these address your specific case.
yindo commented 2026-02-16 18:08:11 -05:00
Author
Owner
Copy Link
Copy Source

@pplmx commented on GitHub (Feb 12, 2026):

hi, @rekram1-node

It seems the opencode models list for NVIDIA NIM is currently backed by a static JSON fixture (packages/opencode/test/tool/fixtures/models-api.json).

There are already signs of divergence:

  • deepseek-ai/deepseek-r1-0528 is still listed but returns HTTP 410 after reaching EOL.
  • stepfun-ai/step-3.5-flash is supported by NVIDIA NIM but does not appear in the list.

Since NVIDIA NIM models can be added or deprecated dynamically, a static model list can easily become stale. This may result in both false positives (deprecated models still listed) and false negatives (newly supported models missing).

Would it make sense to fetch the model list dynamically from the NIM API instead of relying on a static config? Otherwise, the local list may drift from the actual supported models and lead to runtime confusion or failures.

FYI.

@pplmx commented on GitHub (Feb 12, 2026): hi, @rekram1-node It seems the `opencode models` list for NVIDIA NIM is currently backed by a static JSON fixture (`packages/opencode/test/tool/fixtures/models-api.json`). There are already signs of divergence: * `deepseek-ai/deepseek-r1-0528` is still listed but returns HTTP 410 after reaching EOL. * `stepfun-ai/step-3.5-flash` is supported by NVIDIA NIM but does not appear in the list. Since NVIDIA NIM models can be added or deprecated dynamically, a static model list can easily become stale. This may result in both false positives (deprecated models still listed) and false negatives (newly supported models missing). Would it make sense to fetch the model list dynamically from the NIM API instead of relying on a static config? Otherwise, the local list may drift from the actual supported models and lead to runtime confusion or failures. ## FYI. - #13255
Sign in to join this conversation.
No Branch/Tag Specified
Branches Tags
Labels
Clear labels
acp
bad
bug
discussion
docs
documentation
enhancement
extension
good first issue
growth
help-wanted
mcp
model-problem
needs:compliance
nix
opentui
perf
pull-request
Mirrored from GitHub Pull Request
 tui
web
windows
zen
No Label bug zen
Milestone
No items
No Milestone
Projects
Clear projects
No project
Assignees
Clear assignees
yindo
No Assignees
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: anomalyco/opencode#7768
Delete Branch "%!s()"

Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?