DeepSeek V4.1 Flash 1017in opencode

Yes. DeepSeek V4.1 Flash 1017 Uncensored runs in opencode through this endpoint as deepseek-v4.1-flash-1017-uncensored, at $1.25 per million input tokens and $2.50 per million output tokens ($0.30 for cached input). Abliterated DeepSeek for agent loops. DeepSeek V4.1 Flash 1017, the October update, uncensored: a million tokens of context, tool calling, reasoning and vision, with no refusals.

Model id: deepseek-v4.1-flash-1017-uncensored
Build: Abliterated
Context: 1M tokens
Tools / reasoning / images: Yes / Yes / Yes
Per million tokens: $1.25 in / $2.50 out
Base model score: 39 on Artificial Analysis

Where it fits

The October 1017 update of DeepSeek V4.1 Flash, abliterated. It suits long agent loops with many small tool calls, such as running a test suite, reading the failure, editing and running again, where a fast model that never stops to object matters more than the very last point of intelligence.

What to watch for

If you already use DeepSeek in opencode through another provider, give this one its own entry: the ids differ, and this build is the one with the refusals removed.

The config

opencode.json

{
  "$schema": "https://opencode.ai/config.json",
  "provider": {
    "uncensored-opencode": {
      "npm": "@ai-sdk/openai-compatible",
      "name": "Uncensored Opencode",
      "options": {
        "baseURL": "https://uncensoredopencode.com/v1",
        "apiKey": "{env:UNCENSORED_OPENCODE_API_KEY}"
      },
      "models": {
        "deepseek-v4.1-flash-1017-uncensored": {
          "name": "DeepSeek V4.1 Flash 1017 Uncensored",
          "tool_call": true,
          "reasoning": true,
          "attachment": true,
          "limit": { "context": 1048576, "output": 32000 }
        }
      }
    }
  },
  "model": "uncensored-opencode/deepseek-v4.1-flash-1017-uncensored"
}

Your key goes in UNCENSORED_OPENCODE_API_KEY. The opencode page has the setup steps and the block with every model in it.

A first task for it

shell

opencode run --model uncensored-opencode/deepseek-v4.1-flash-1017-uncensored "Run the test suite, fix the first failure, and repeat until it passes"

The score above is the base model's, from the Artificial Analysis Intelligence Index v4.3.2. We do not claim this build scores the same.

Next to it on the line: GLM 5.3 Flash, Kimi K3.1 Pro, GLM 5.3, MiMo V2.6 Flash, Qwen3.8 27B, GLM 5.3 Flash Cyber.

Bring your agent.

Get an API key Set up opencode See the models