IBM Granite 4.2 8B
IBM · United States · Apache 2.0
Commercial use: Yes — free for commercial use
Apache 2.0. IBM publishes provenance and governance documentation alongside the weights, which is unusual and useful in a procurement pack.
A small, permissively licensed model from IBM that ships with documentation of what it was trained on.
Why it should matter to you
Most open models cannot tell you where their training data came from. That question is live in litigation and increasingly appears in enterprise procurement. Granite is one of the few families with an answer, and the answer is written down.
How it connects to our work
Where a client has real provenance exposure — a product that ships generated output, or a regulated buyer asking the question — we recommend a provenance-documented model even at a capability cost. It is a risk decision and it should be made on purpose.
Advantages
- Apache 2.0 from a vendor whose enterprise paperwork your legal team already recognises
- Documented training-data governance — a real differentiator when someone asks what it was trained on
- Official GGUF, FP8, NVFP4 and MXFP4 builds published by IBM
- Small enough for an 8 GB card
Disadvantages
- Not competitive with Qwen3.8-27B on raw capability
- Modest community fine-tune ecosystem
- Conservative outputs; less useful for open-ended generation
Reach for it when
Regulated enterprises where training-data provenance is a procurement question and IBM as a counterparty carries weight.
Where it falls down
Anything needing frontier reasoning. This is a governance choice, not a capability choice, and it should be made deliberately.
Running it
8 GB GPU or 16 GB Mac at Q4_K_M. See the hardware sizing tables for how that maps to specific chips and cards, and the quantisation guide for what you give up at each bit width.
Which cloud instance actually runs this, and what it costs
At Q4_K_M the weights are about 5 GB, so 20 catalogued instances fit with room for context, of which the 8 best value are shown. Cheapest per token is Hetzner GEX45 (~$0.60 per million output tokens at an estimated 134 tok/s). Fastest on a single card is Azure NCads H100 v5 at roughly 670 tok/s for $2.89 per million tokens, which is the speed-versus-cost trade in one line.
| Instance | GPU | VRAM | Bandwidth | Est. tok/s | $/hr | $/M tokens |
|---|---|---|---|---|---|---|
| Hetzner GEX45 | RTX PRO 4000 Blackwell | 24 GB | 672 GB/s | ~134 | $0.290 | $0.60 |
| RunPod A100 PCIe | A100 80GB | 80 GB | 1935 GB/s | ~387 | $1.190 | $0.85 |
| Hetzner GEX131 | RTX PRO 6000 Blackwell Max-Q | 96 GB | 1792 GB/s | ~358 | $1.420 | $1.10 |
| RunPod L40S | L40S | 48 GB | 864 GB/s | ~173 | $0.790 | $1.27 |
| RunPod H100 PCIe | H100 PCIe | 80 GB | 2000 GB/s | ~400 | $1.990 | $1.38 |
| Lambda A100 40GB | A100 40GB | 40 GB | 1555 GB/s | ~311 | $1.990 | $1.78 |
| Lambda H100 PCIe | H100 PCIe | 80 GB | 2000 GB/s | ~400 | $3.290 | $2.28 |
| AWS g5.xlarge | A10G | 24 GB | 600 GB/s | ~120 | $1.006 | $2.33 |
Token generation on a single request is memory-bandwidth-bound, not compute-bound: the GPU must read every active weight from memory for each token it emits. So the ceiling is roughly GPU memory bandwidth divided by the size of the active weights. A 16 GB model on a 300 GB/s L4 tops out near 19 tokens/sec; the same model on a 2039 GB/s A100 tops out near 127. Figures on model pages are that arithmetic, computed from the bandwidth and weight-size columns here. They are a ceiling for one request at batch size 1, before serving overhead, so treat them as an upper bound for comparing instances rather than a benchmark. Batching raises total throughput well above this and does not raise per-request speed.
For identical silicon, the hyperscalers charge roughly 2 to 5 times what the specialist GPU clouds charge. An A100 80GB is about $3.43/GPU-hr on AWS, $3.67 on Azure and $5.03 on Google Cloud, against $1.19-1.39 on RunPod and $1.99 on Lambda. That spread is the single largest cost lever in a self-hosted inference budget, and it is larger than any saving from picking a smaller model.
Three costs that are absent from every headline rate: egress, storage for the weights, and idle time. A 70B model at Q4 is tens of gigabytes that must be pulled to the instance on every cold start, and an endpoint billed hourly is billed while it sits idle waiting for a request. Rates read 2026-09-06; check each provider's live pricing on the directory hub.
Free tiers carrying this model
| Provider | Type | The catch | Live limits |
|---|---|---|---|
| NVIDIA NIM | Inference host | It is a credit grant, not a permanent free tier. When the credits run out you are on a paid plan or self-hosting the container. | check → |
| OpenRouter | Aggregator | Free-pool membership changes without notice — a model you built on can stop being free, and the list above will drift. Data passes through OpenRouter and then the upstream provider, so check the privacy and routing settings before sending anything sensitive. | check → |
Jurisdiction: United States
Best raw capability and the deepest tooling ecosystem. For EU personal data you are relying on a transfer framework rather than on data never leaving the bloc, so check whether your DPA and your customers accept that.
Watch for: Enterprise API tiers usually promise no training on your data; consumer tiers and free tiers frequently do not. The free tier is where this bites.
Frequently asked
Can I use IBM Granite 4.2 8B commercially?
Apache 2.0. IBM publishes provenance and governance documentation alongside the weights, which is unusual and useful in a procurement pack.
What hardware do I need to run IBM Granite 4.2 8B?
8 GB GPU or 16 GB Mac at Q4_K_M. Weights alone are roughly 5 GB at Q4_K_M and 9 GB at Q8_0. Add KV cache on top of that, which grows with your context length.
What licence is IBM Granite 4.2 8B released under?
Apache 2.0. Full commercial use, modification and redistribution. Patent grant included. The most permissive licence in common use for open-weight models.
Where can I use IBM Granite 4.2 8B for free?
Free tiers carrying it include NVIDIA NIM, OpenRouter. Limits differ per provider and change often, so check each provider's own limits page. You can also self-host the weights, which has no rate limit at all.
Similar models
gpt-oss-120b
80 GB GPU, or 96 GB+ unified memory.
gpt-oss-20b
16 GB GPU or Mac. A strong default for local agents.
Gemma 4 31B
24 GB GPU at Q4_K_M, or a 32 GB Mac.
Fara-7B
8 GB GPU or any 16 GB Mac.
Wiring IBM Granite 4.2 8B into something real?
We build the evaluation harness, the failover and the cost ceilings around a model like this, so it survives contact with production.
Maps to AI agent development, AI automation development and AI strategy and consulting. Or see it working: our case studies.