Written by Jakub Rusinowski · Last updated September 6, 2026
Model library → VibeThinker → VibeThinker 1.5B
The original VibeThinker, fine-tuned from Qwen2.5-Math-1.5B. Reports 80.3 / 74.4 / 50.4 on AIME24, AIME25 and HMMT25 — above DeepSeek R1 on those three at roughly 1/400th the parameter count. Under 2 GB at Q4_K_M, so it runs on a phone.
VibeThinker 1.5B needs about 2 GB of VRAM at Q4_K_M — quantized weights plus framework overhead, before any KV cache. On Apple Silicon that figure comes out of unified memory.
| Parameters | 1.54 Billion |
| Context window | 131,072 |
| Architecture | Dense Transformer (Qwen2.5-Math-1.5B base) |
| Provider | WeiboAI |
| Licence | MIT |
| Specified at | Q4_K_M |
| System RAM | 4 GB |
| Record updated | 2026-09-06 |
MIT — commercial use permitted. Commercial use permitted. No usage restrictions beyond attribution.
Modelled on a reference NVIDIA RTX 4090 (24 GB), with no KV cache (this record has no published architecture). Speed figures are ESTIMATES from the memory-bandwidth roofline described on the methodology page, not benchmarks we ran — rows marked measured come from published or reader-submitted runs. VRAM here includes the KV cache, so it reads higher than the headline figure above, which does not.
| Quant | Weights | VRAM needed | Est. speed | Fit on 24 GB |
|---|---|---|---|---|
| Q2_K | 0.5 GB | 1.3 GB | ~298 tok/s (est.) | Fits comfortably |
| Q3_K_M | 0.7 GB | 1.5 GB | ~281 tok/s (est.) | Fits comfortably |
| Q4_K_M | 0.9 GB | 1.7 GB | ~255 tok/s (est.) | Fits comfortably |
| Q5_K_M | 1.1 GB | 1.9 GB | ~242 tok/s (est.) | Fits comfortably |
| Q6_K | 1.3 GB | 2.1 GB | ~229 tok/s (est.) | Fits comfortably |
| Q8_0 | 1.6 GB | 2.4 GB | ~206 tok/s (est.) | Fits comfortably |
| F16 | 3.1 GB | 3.9 GB | ~148 tok/s (est.) | Fits comfortably |
Want the memory numbers alone, at every quantization level and your own context length? Use the VibeThinker 1.5B VRAM calculator.
or compare on Vast.ai from $0.35/hr (typical low · varies)
As an Amazon Associate we earn from qualifying purchases. Cloud GPU links are referral links — we may earn a commission at no extra cost to you.
The cheapest catalogued GPU that runs VibeThinker 1.5B is the Intel Arc B570 (10 GB).
Install Ollama, then run:
ollama run vibethinkerWeights on Hugging Face: WeiboAI/VibeThinker-1.5B.
Quality scores as published by the model's authors or an independent evaluator — not throughput, and not measured by us.
| Benchmark | Score | Provenance |
|---|---|---|
| AIME 2024 | 80.3 / 100 % | vendor-claimed · https://github.com/WeiboAI/VibeThinker |
| AIME 2025 | 74.4 / 100 % | vendor-claimed · https://github.com/WeiboAI/VibeThinker |
Best for: reasoning, math, edge devices, research.
← All VibeThinker models | VRAM calculator | Check your own hardware