fwizzer1/Fwizzer-R1-3B-RU

Update chat_template.jinja with mandatory <think> prefill

Fix Fwizzer-R1.preset.json encoding, system prompt, and reasoning parser

Rename GGUF files to include -RU- prefix

Update Fwizzer-R1-3B-Max.gguf with full weights and embedded reasoning template

Update Fwizzer-R1-3B-Speed.gguf with embedded reasoning system prompt and enable_thinking

Synchronize verified Fwizzer-R1-3B-Speed GGUF with embedded think prefill

Add official Fwizzer-R1 LM Studio preset

Update chat_template.jinja with Fwizzer identity and <think> prefill

Update Fwizzer-R1-3B-Speed.gguf with clean explicit <think> generation

Fix chat template so model outputs <think> explicitly for UI reasoning parsers

Update README with Speed and Max filenames

Update model.yaml with Speed and Max filenames

Rename model files to Fwizzer-R1-3B-Max.gguf and Fwizzer-R1-3B-Speed.gguf

Rename Q8_0 file to Fwizzer-R1-3B.Q8_0.gguf

Inject baked Fwizzer-R1 system prompt directly into GGUF binary header

Update README with full documentation and reasoning tags

Add model.yaml with reasoning: true, thinking: true, deepseek format

Bake exact Fwizzer-R1 system prompt into chat_template.jinja

Clean up Fwizzer-R1-3B-Instruct-2512.Q4_K_M.gguf

Update model.yaml: Standard UI mode for Fwizzer-R1-3B

Update chat_template.jinja: Name Fwizzer-R1-3B without forced thinking

Add clean Fwizzer-R1-3B.Q4_K_M.gguf (Name: Fwizzer-R1-3B, standard chat without forced thinking)

Remove old Ministral-named Q4_K_M duplicate

Add Fwizzer-R1-3B-Instruct-2512.Q4_K_M.gguf with baked metadata

Bake Fwizzer-R1-3B-RU name and reasoning chat_template directly into GGUF binary

Enable 1M Context (1,048,576 tokens) with YaRN RoPE scaling

Enable 1M Context (1048576 tokens) in model.yaml

Update README: generalized code generation description

Publish professional English README with benchmarks, GGUF table and quickstart

Set reasoning: true, thinking: true and deepseek format in model.yaml

Add native deepseek reasoning token format to tokenizer_config.json

Enable native reasoning and thinking capability flags in config.json

Update README with official reasoning tags and description

Add LM Studio model.yaml with reasoning: true

Set default_system_prompt to Fwizzer-R1-3B-RU in tokenizer_config.json

Bake Fwizzer-R1-3B-RU identity and system prompt

Bake reasoning chat_template and system prompt into tokenizer_config.json

Set built-in default reasoning system prompt and auto <think> generation

Add README

Trained with Unsloth - config

Trained with Unsloth

Trained with Unsloth

Upload model trained with Unsloth

Upload model trained with Unsloth

Upload model trained with Unsloth

Upload model trained with Unsloth

Upload model trained with Unsloth

Upload model trained with Unsloth

Upload model trained with Unsloth

Upload README.md with huggingface_hub