NVIDIA API models hang indefinitely with "Aborted process" - all models affected (DeepSeek V4, Llama, Mistral, Minimax)

[alexbayov]

Description

Experiencing the same issue with NVIDIA API models via https://integrate.api.nvidia.com/v1

Configuration: All NVIDIA models (deepseek-ai/deepseek-v4-pro, meta/llama-3.3-70b-instruct, mistralai/mistral-large, minimaxai/minimax-m2.7) hang with "Размышляю" status and eventually fail with "Aborted process" error.

Environment:

• OpenCode Version: 25
• OS: macOS 26.2 (25C56)

Logs:
ERROR 2026-04-26T16:45:20 +230498ms service=session.processor error=Aborted process

Direct curl requests to NVIDIA API work correctly, suggesting streaming response handling issue.

Plugins

No response

OpenCode version

No response

Steps to reproduce

No response

Screenshot and/or share link

No response

Operating System

No response

Terminal

No response

[github-actions]

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

• Nvidia NIM API hangs for DeepSeek v4 reasoning models without chat_template_kwargs #24264: Nvidia NIM API hangs for DeepSeek v4 reasoning models — same provider (NVIDIA NIM), same models (DeepSeek V4), and same symptom (hangs indefinitely). That issue identifies a specific root cause: NVIDIA NIM requires chat_template_kwargs in the payload for DeepSeek v4 reasoning models, which OpenCode currently doesn't inject. Your non-DeepSeek models (Llama, Mistral, Minimax) hanging may be a separate issue, but the DeepSeek V4 hang is likely the same.

[github-actions]

To stay organized issues are automatically closed after 60 days of no activity. If the issue is still relevant please open a new one.