🍓 Strawberry-1.1

Strawberry-1.1 is a fine-tuned version of artindnr/strawberry-1, further trained on a hybrid English-Persian dataset to strengthen dual-linguality reasoning — the model's ability to reason fluently in English and Farsi at once, without the two languages interfering with each other mid-thought.

Model Details

• Base model: artindnr/strawberry-1 (itself fine-tuned from openai/gpt-oss-20b, 21B parameters)
• Architecture: gpt_oss
• Fine-tuned by: artindnr
• License: Apache 2.0
• Languages: Farsi (Persian), English, and multilingual reasoning support
• Model type: Causal decoder-only language model with reasoning ("thinking") traces

What's New

Strawberry-1 was the first ~20B-class open model to produce high-quality Farsi reasoning traces, but it could run into language conflict: when a prompt pulled on both English and Farsi at once, its reasoning could become inconsistent or unstable as the two languages competed within the same trace.

Strawberry-1.1 was fine-tuned specifically to resolve this. Relative to Strawberry-1, it:

• Reasons across English and Farsi within the same trace without the languages conflicting, where Strawberry-1 would show that conflict
• Preserves the reasoning-language control and general instruction-following ability inherited from Strawberry-1
• Builds on the same Harmony chat template and usage pattern as Strawberry-1, so it's a drop-in upgrade

Training

Strawberry-1.1 was produced by further fine-tuning artindnr/strawberry-1 on a hybrid English-Persian dataset.

Training Data

Strawberry-1.1 was trained on a hybrid English-Persian dataset drawn from the Thinking Datasets collection, the same collection used to train the base Strawberry-1 model.

How to Use

Strawberry-1.1 uses the gpt-oss chat template (Harmony format) inherited from the base model, so it works with 🤗 Transformers.

Installation

pip install torch --index-url https://download.pytorch.org/whl/cu128
pip install "trl>=0.20.0" "peft>=0.17.0" "transformers>=4.55.0" "kernels>=0.12.0"

This has been verified to work with:

Generation

import torch
from transformers import AutoModelForCausalLM, AutoTokenizer

MODEL_ID = "artindnr/strawberry-1.1"

tokenizer = AutoTokenizer.from_pretrained(MODEL_ID)
model = AutoModelForCausalLM.from_pretrained(
    MODEL_ID,
    torch_dtype=torch.bfloat16,
    device_map="auto",
)

REASONING_LANGUAGE = "English"  # e.g. "English", "Farsi", "Persian"
SYSTEM_PROMPT = f"reasoning language: {REASONING_LANGUAGE}"
USER_PROMPT = "تو کی هستی و اسمت چیه؟"

messages = [
    {"role": "system", "content": SYSTEM_PROMPT},
    {"role": "user", "content": USER_PROMPT},
]

inputs = tokenizer.apply_chat_template(
    messages,
    add_generation_prompt=True,
    tokenize=True,
    return_dict=True,
    return_tensors="pt",
).to(model.device)

outputs = model.generate(
    **inputs,
    max_new_tokens=512,
    temperature=0.6,
    do_sample=True,
)

print(tokenizer.decode(outputs[0]))

This prints the full Harmony-formatted output, including the analysis (reasoning) and final (answer) channels and their special tokens. To get just the plain-text final answer, decode with skip_special_tokens=True and parse out the final channel, or use tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:], skip_special_tokens=True) to only decode the newly generated tokens.

Reasoning in a specific language

Set reasoning language: <Language> as the system message content to control the language of the reasoning trace (the analysis channel), independent of the language the user writes in. For example, setting REASONING_LANGUAGE = "Farsi" will produce a Farsi reasoning trace even for a prompt in another language. Strawberry-1.1's fine-tuning specifically targeted keeping this stable when a prompt draws on both languages at once.

Note that the model's default chat template also auto-populates a Harmony-format preamble (identity, knowledge cutoff, current date, reasoning effort, valid channels) ahead of your system/developer message — you don't need to set these yourself.

Intended Use

Strawberry-1.1 is intended for:

• Research and experimentation on multilingual and Farsi-language reasoning, particularly cross-lingual (English-Farsi) reasoning
• Building Farsi-language assistants, tutoring tools, and reasoning-heavy applications
• General-purpose multilingual chain-of-thought tasks

Limitations

• Farsi reasoning quality, while a focus of this fine-tune, may still occasionally mix in English tokens or phrasing, especially for highly technical topics.
• As with any fine-tune, Strawberry-1.1 inherits the general capabilities and limitations of the strawberry-1 and gpt-oss-20b base models, including the possibility of hallucinated facts and reasoning errors.
• No formal safety fine-tuning beyond what is inherited from the base models has been applied; use appropriate safeguards in production settings.

License

This model is released under the Apache 2.0 license, consistent with the base strawberry-1 and gpt-oss-20b models.

Citation

If you use Strawberry-1.1 in your work, please cite:

@misc{strawberry11,
  title  = {Strawberry-1.1: Resolving English-Farsi Reasoning Conflict in a Fine-tuned GPT-OSS-20B Model},
  author = {artindnr},
  year   = {2026},
  url    = {https://huggingface.co/artindnr/strawberry-1.1}
}

Acknowledgements

Built on top of artindnr/strawberry-1 and, in turn, openai/gpt-oss-20b, using the Thinking Datasets collection.