GPT-OSS 20B fails with Continue VSCode extension

[softdev-rsuza]

Describe the bug

Using guide from https://docs.openvino.ai/2026/model-server/ovms_demos_code_completion_vsc.html, GPT-OSS 20B with OVMS and Continue VSCode extension results in invalid gptoss reasoning output handling.

Continue extension returns:

400 INVALID_ARGUMENT: CalculatorGraph::Run() failed:
Calculator::Process() for node "LLMExecutor" failed:
You have passed a message containing <|channel|> tags in the content field.
Instead of doing this, you should pass analysis messages
(the string between '<|message|>' and '<|end|>')
in the 'reasoning_content' field,
and final messages in the 'content' field.

---

To Reproduce

1. Follow official documentation

Followed the official guide:
https://docs.openvino.ai/2026/model-server/ovms_demos_code_completion_vsc.html

2. OVMS launch command

ovms --model_repository_path ~/models \
     --source_model OpenVINO/gpt-oss-20b-int4-ov \
     --task text_generation \
     --target_device GPU \
     --tool_parser gptoss \
     --reasoning_parser gptoss \
     --rest_port 8000 \
     --model_name gpt-oss-20b

3. Continue config

- name: OVMS gpt-oss-20b
  provider: openai
  model: gpt-oss-20b
  apiKey: unused
  apiBase: http://localhost:8000/v3
  roles:
    - chat
    - edit
    - apply
  capabilities:
    - tool_use

4. Trigger issue

Use Continue chat in VSCode with GPT-OSS model.

Example prompt:

Create a simple hello world PHP example.

OVMS responds with:

400 INVALID_ARGUMENT: CalculatorGraph::Run() failed:
Calculator::Process() for node "LLMExecutor" failed:
You have passed a message containing <|channel|> tags in the content field.
Instead of doing this, you should pass analysis messages
(the string between '<|message|>' and '<|end|>')
in the 'reasoning_content' field,
and final messages in the 'content' field.

---

Expected behavior

Expected GPT-OSS responses to work correctly with Continue/OpenAI-compatible chat clients when following the official OVMS tutorial.

---

Logs

OVMS Version

OpenVINO Model Server 2026.2.0.5e9dcfc46
OpenVINO backend 2026.2.0-21903-52ddc073857-releases/2026/2
OpenVINO GenAI backend 2026.2.0.0-3121-adf73e80e66
Bazel build flags: --strip=always --config=mp_on_py_on --//:distro=ubuntu

---

Configuration

Not used.

3. Hardware

• Intel Ultra 7 258v, Intel Arc 140v
• 32 GB DDR5
• Ubuntu 24.04