Skip to content

Instantly share code, notes, and snippets.

View maddes8cht's full-sized avatar

maddes8cht maddes8cht

View GitHub Profile

Best Current OpenCode Models

Last Updated: August 29, 2026

This ranking blends the free models exposed by opencode models opencode with strong reference models (DeepSeek-V4 Pro/Flash, Qwen3.8-27B) that are not bundled by default but are top-tier candidates for OpenCode workflows via API or local serving.

Rank Model Name Type/Provider Context Window Key Strengths OpenCode Suitability
1 DeepSeek V4 Pro Cloud/Free (open weights, MIT) 1M tokens Frontier reasoning & coding, SOTA open agentic coding (SWE-bench >96%), three reasoning modes (Non-think/Think High/Think Max), ultra-efficient 1M context (CSA/HCA) Best-in-class for hard agentic/terminal coding, repo-scale and multi-step tasks; natively integrated with OpenCode
2 DeepSeek V4 Flash Cloud/Free (open weights, MIT) 1M tokens Reasoning near Pro, 13B-active MoE efficiency, very low cost/latency, 1M context default E
@maddes8cht
maddes8cht / best-current-opencode-models.md
Created August 22, 2026 13:53
Ranking & Steckbriefe der aktuellen kostenlosen OpenCode-Zen-Modelle (Stand: 22.08.2026)

Best Current OpenCode Models (Free Tier)

Dokument erstellt am: Samstag, 22. August 2026 (Modellliste erhoben via opencode models opencode; Specs aus models.dev, offizieller OpenCode-Zen-Doku und Herstellerquellen, Stand dieses Tages.)

OpenCode Zen listet aktuell sieben kostenlose Modelle (Input/Output/Cached Read = $0, zeitlich begrenzt). Die Modell-IDs rotieren – der Live-Stand ist immer über opencode models opencode bzw. https://opencode.ai/zen/v1/models abrufbar.

Transparenzhinweis: Dieses Dokument wurde von ox-alpha verfasst – dem Modell hinter x-preview-f-free. Die Einordnung folgt ausschließlich dokumentierter Evidenz (Benchmarks, Specs, Praxisberichte), nicht Eigeninteresse.


@maddes8cht
maddes8cht / generate-cmd.md
Last active February 1, 2026 11:00
Python Gist

Generate-cmd.py

This script does the following:

  • takes 2 parameters:
  • one param determines a python script somewhere on your system
  • other param determines a path on your system where the resulting .cmd file will be placed
  • if any param is missing, a file / path selection dialog will be opened to select one
  • python file will be run with parameter -h. Output gets captured and is included with leading commets in the resulting .cmd file
  • a .cmd file is placed in the directory path with the same nam as the .py file, but with extension .cmd
  • The cmd file will call the .py file at its given location

Makefiles Evolution

My makefiles had some evolution over time. Here is a list of Makefiles that did a good job for me.

M1 is a simple makefile M2 is a makefile with much better automatic checking of object files

M4 is a bit too advanced in having too much possible targets that are unnecessary, esp. the run targets. there is no use of a run-debug target, as the debugger needs to be hooked by the calling process to work, which is not possible in a makefile.

M5 is a makefile that is a bit less complex, leaving out unecessary targets.

@maddes8cht
maddes8cht / MSys 2.md
Last active August 25, 2024 10:58
MSys2 setup

MSys 2

installation guide, hints and cooperation with VS Code

  • nmake.bat hat den Zweck, aufrufe von nmake nach make umzuleiten, wenn nmake im Pfad nicht vorhanden ist
@maddes8cht
maddes8cht / os.walk-sample.py
Last active January 22, 2023 15:31
Python schnipsel
import os
print ('\n #######\n')
print (os.getcwd)
os.chdir('C:\\Users\\WaWiAdm\\Documents\\MBa\\PalvinBabs\\Babs 2018')
print (os.getcwd)
print (os.listdir())
print ('\n ********\n')