Model comparison
qwen2.5-3b-grpo-gsm8k vs GPT-5
Capability radar, value bars, and side-by-side posture.
qwen2.5-3b-grpo-gsm8k
Not scored
value score
GPT-5
94
value score
qwen2.5-3b-grpo-gsm8k ctx
—
GPT-5 ctx
400K
Capability radar
Value · context · multimodal · openness · speed posture
Value head-to-head
- GPT-594
Attribute tape
| Field | qwen2.5-3b-grpo-gsm8k | GPT-5 |
|---|---|---|
| Developer | sarimahsan101 | OpenAI |
| Context | — | 400K |
| Modalities | Text, Code, Image, Audio | |
| Openness | — | Proprietary Model |
| Speed | — | Fast |
| Price | — | Mid-High |
| Value score | — | 94 |
| Summary | A 3-billion-parameter Qwen2.5 language model fine-tuned with Group Relative Policy Optimization (GRPO) on the GSM8K grade-school math benchmark. It is designed to generate step-by-step arithmetic and word-problem reasoning on modest hardware. | OpenAI frontier multimodal model. |