Granite 4.0 H-Small 32B-A9Bvsgpt-oss 20B
Complete comparison between Granite 4.0 H-Small 32B-A9B (32B parameters, IBM) and gpt-oss 20B (21B, OpenAI). VRAM required by quantization, tokens per second measured on 4 reference GPUs, verdict by use case, license, installation commands. All figures are calculated from catalog data—no copy-pasting between pages.
In brief
| Feature | Granite 4.0 H-Small 32B-A9B | gpt-oss 20B |
|---|---|---|
| Settings | 32B | 21B |
| Family | Granite | gpt-oss |
| Author | IBM | OpenAI |
| Origin | US | US |
| License | Apache 2.0 | Apache 2.0 |
| Context | 128,000 tokens | 128,000 tokens |
| Output | October 2025 | April 2025 |
Memory footprint
Approximate VRAM required for inference with an average context window. The winner (in green) is the model that uses less — smaller models have the edge.
| Quantization | Granite 4.0 H-Small 32B-A9B | gpt-oss 20B |
|---|---|---|
| Q4_K_M (lightweight) | 19 GB | 13 GB |
| Q5_K_M (balanced) | 23 GB | 16 GB |
| Q8 (near-lossless) | 35 GB | 23 GB |
| FP16 (max quality) | 64 GB | 42 GB |
| CPU-only RAM | 32 GB | 18 GB |
Estimated speed (tokens/second)
Estimates based on the best quantization that can fit on each GPU. Actual figures depend on the prompt, context, and engine (llama.cpp, vLLM, MLX). Methodology.
| Reference GPU | Granite 4.0 H-Small 32B-A9B | gpt-oss 20B |
|---|---|---|
| RTX 4090 (24 GB) | 75 tok/s · Q5_K_M | 130 tok/s · Q8 |
| RTX 4080 (16 GB) | ✗ too heavy | 55 tok/s · Q5_K_M |
| RTX 3060 12GB (12 GB) | ✗ too heavy | ✗ too heavy |
| Apple M4 Pro (48 GB) (36 GB) | 30 tok/s · Q8 | 55 tok/s · Q8 |
Verdict by use case
For each common use case, we indicate which of the two is better suited based on its tags, size, and specialization.
Your private, free ChatGPT on your machine in 1 hour — LM Studio, Ollama, Open WebUI, your documents, no cloud.
- Lifetime online access
- PDF + files
- Refunded within 30 days
Strengths and weaknesses
Granite 4.0 H-Small 32B-A9B
Mamba-2 + MoE 32B/9B hybrid. ~70% less RAM in long contexts. Apache 2.0.
- 70% less RAM in long-context workloads
- Apache 2.0
- Enterprise-ready
- Recent llama.cpp support required
Installation
gpt-oss 20B
Little brother of gpt-oss 120B. 21B/3.6B active. Matches o3-mini on a laptop.
- MIT License
- 13 GB VRAM Q4
- OpenAI quality in an accessible format
- 128k context
- MoE — more VRAM than an equivalent dense model
- Fewer fine-tunes than Llama/Qwen
Installation
Frequently asked questions
Which model runs best on RTX 4090 (24 GB): Granite 4.0 H-Small 32B-A9B or gpt-oss 20B?
On an RTX 4090, Granite 4.0 H-Small 32B-A9B runs at Q5_K_M (~75 tok/s), gpt-oss 20B in Q8 (~130 tok/s). In pure throughput terms, gpt-oss 20B wins. See the configurator to test your exact GPU.
Which uses the least VRAM?
In Q4_K_M, gpt-oss 20B fits in 13 GB versus 19 GB for its rival—a 6 GB gap, significant if you are targeting an RTX 3060 or an 8 GB 4060 Ti.
Granite 4.0 H-Small 32B-A9B or gpt-oss 20B for coding?
gpt-oss 20B is the best fit (code-specialized, top HumanEval scores). For an IDE copilot, pair it with Continue or Aider.
Can these models be used in commercial production?
Granite 4.0 H-Small 32B-A9B is under Apache 2.0 — free for commercial use. gpt-oss 20B is under Apache 2.0 — also free to use. For SaaS, prefer Apache 2.0 or MIT.
Which one should you choose in 2026?
Depends on your primary constraint. Smaller / faster: gpt-oss 20B (21B). More capable: Granite 4.0 H-Small 32B-A9B (32B). If you're unsure, run the configurator with your GPU and use case — it will decide based on both.
Alternatives to consider
If neither works for you, here are some neighboring models you could explore.