GLM-5.3 is live. Run Z.ai's latest model on Friendli Model APIs. Try it today ➜
karimmashaly
Available on FriendliAI
Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Run this model inference with full control and performance in your environment.
Model Details
Model Provider
karimmashaly
Model Tree
Qwen/Qwen2.5-1.5B
Input Modalities
Output Modalities
Supported Functionality
On 437 held-out prompts, the most recent 15% of dates for each ticker, the model produced one answer per prompt:
Most answers still run until the completion cap. This is a research checkpoint, not a trading system.
python
from transformers import AutoModelForCausalLM, AutoTokenizer model_id = "karimmashaly/egx-qwen2.5-1.5b-grpo"tokenizer = AutoTokenizer.from_pretrained(model_id)model = AutoModelForCausalLM.from_pretrained(model_id, torch_dtype="auto", device_map="auto") messages = [ { "role": "system", "content": ( "You are an Egyptian Stock Exchange analyst. Reason from the supplied " "price history and news only. Write in English. Do not invent numbers " "that are not supported by the context. " "Put the reasoning inside <think></think> and the return class inside " "<answer></answer>. The answer must be exactly one of: " "Strong Bullish, Slight Bullish, Neutral, Slight Bearish, Strong Bearish." ), }, { "role": "user", "content": ( "Ticker: COMI (Commercial International Bank)\n" "Analysis date T: 2026-01-15\n\n" "Recent price history (sessions strictly before T):\n...\n\n" "News from T-7 through T:\n...\n\n" "Task: Predict the return class of this stock over the next 5 trading days. " "The allowed labels are: Strong Bullish, Slight Bullish, Neutral, Slight Bearish, Strong Bearish. " "Reply in exactly this form and nothing else: " "<think>your reasoning</think><answer>Label</answer>. " "The answer tag must contain one allowed label and no other text." ), },]prompt = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)inputs = tokenizer(prompt, return_tensors="pt").to(model.device)output = model.generate(**inputs, max_new_tokens=1024)print(tokenizer.decode(output[0][inputs["input_ids"].shape[-1]:], skip_special_tokens=True))from transformers import AutoModelForCausalLM, AutoTokenizer model_id = "karimmashaly/egx-qwen2.5-1.5b-grpo"tokenizer = AutoTokenizer.from_pretrained(model_id)model = AutoModelForCausalLM.from_pretrained(model_id, torch_dtype="auto", device_map="auto") messages = [ { "role": "system", "content": ( "You are an Egyptian Stock Exchange analyst. Reason from the supplied " "price history and news only. Write in English. Do not invent numbers " "that are not supported by the context. " "Put the reasoning inside <think></think> and the return class inside " "<answer></answer>. The answer must be exactly one of: " "Strong Bullish, Slight Bullish, Neutral, Slight Bearish, Strong Bearish." ), }, { "role": "user", "content": ( "Ticker: COMI (Commercial International Bank)\n" "Analysis date T: 2026-01-15\n\n" "Recent price history (sessions strictly before T):\n...\n\n" "News from T-7 through T:\n...\n\n" "Task: Predict the return class of this stock over the next 5 trading days. " "The allowed labels are: Strong Bullish, Slight Bullish, Neutral, Slight Bearish, Strong Bearish. " "Reply in exactly this form and nothing else: " "<think>your reasoning</think><answer>Label</answer>. " "The answer tag must contain one allowed label and no other text." ), },]prompt = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)inputs = tokenizer(prompt, return_tensors="pt").to(model.device)output = model.generate(**inputs, max_new_tokens=1024)print(tokenizer.decode(output[0][inputs["input_ids"].shape[-1]:], skip_special_tokens=True))