kforge

Presets Inventory

This document is a snapshot of the current built-in model presets in KForge. It reflects the state at the time of Phase 3.12 and is intended as a reference for documentation, maintenance, and future remote presets work.

Cost is represented by color labels, and usage is represented separately.

Legend:

Usage:

Claude

Model Cost Usage Notes
claude-opus-4-5 🔴 Paid Heavy Highest capability; use sparingly
claude-sonnet-4-5 🟡 Paid Main Balanced default for dev + writing
claude-haiku-4-5 🟢 Paid Sandbox Cheap + fast; small tasks

✅ No free Claude — consistent with provider pricing.

Custom (OpenAI-compatible endpoints)

Model Cost Usage Notes
deepseek-chat 🔵 Free Sandbox Provider-dependent
llama-3.1-8b-instant 🔵 Free Sandbox Provider-dependent
mistral-medium-latest 🟡 Paid Main Paid workhorse
mistral-small-latest 🟢 Paid Sandbox Low-cost paid
openai/gpt-4o-mini 🟢 Paid Sandbox Cheap paid default

⚪ “Free” here is endpoint-dependent and not guaranteed.

DeepSeek

Model Cost Usage Notes
deepseek-reasoner 🟡 Paid Main Stronger reasoning; slower/costlier
deepseek-chat 🟢 Paid Sandbox Cheap general chat

Gemini (preview-heavy)

Model Cost Usage Notes
gemini-3-pro-preview 🔴 Paid Heavy Preview; high capability
gemini-2.5-pro 🟡 Paid Main Strong reasoning
gemini-3-flash-preview 🟡 Paid Main Preview; may change
gemini-2.5-flash 🟢 Paid Sandbox Fast
gemini-2.5-flash-lite 🟢 Paid Sandbox Fast + cheap

⚠ Preview models are subject to change or removal.

Groq

Model Cost Usage Notes
llama-3.3-70b-versatile 🟡 Paid Main Large + fast
llama-3.1-8b-instant 🟢 Paid Sandbox Very fast

Mistral (Hosted)

Model Cost Usage Notes
codestral-latest 🟡 Paid Main Coding-focused
mistral-small-latest 🟢 Paid Sandbox General starter

Ollama (Local)

Model Cost Usage Notes
codellama:13b 🔴 Paid* Heavy Local compute cost
deepseek-coder:6.7b 🟡 Paid* Main Local
llama3.1:8b 🟡 Paid* Main Local default
qwen2.5-coder:7b 🟡 Paid* Main Local
mistral:7b 🟢 Paid* Sandbox Fast
qwen2.5-coder:1.5b 🟢 Paid* Sandbox Very fast

* Paid refers to local hardware/energy cost, not API billing.

OpenAI

Model Cost Usage Notes
gpt-4.1-mini 🟡 Paid Main Day-to-day
gpt-5-mini 🟢 Paid Sandbox Cheap paid testing

OpenRouter (highly volatile)

Model Cost Usage Notes
meta-llama/llama-3.3-70b-instruct:free 🔵 Free Sandbox Rotating / rate-limited
mistralai/devstral-2512:free 🔵 Free Sandbox Rotating
qwen/qwen3-coder:free 🔵 Free Sandbox Availability may change
xiaomi/mimo-v2-flash:free 🔵 Free Sandbox Availability may change

⚠ OpenRouter free models are not guaranteed and may disappear without notice.

Duplicate Models (Intentional Overlap)

Model Providers
deepseek-chat Custom, DeepSeek
llama-3.1-8b-instant Custom, Groq
mistral-small-latest Custom, Mistral

Overlaps are intentional and reflect different tradeoffs (cost, routing, availability, privacy).

Volatility Watchlist

High risk:

Medium risk:

Low risk: