Granite 4.2 30B
Granite 4.2 30B needs 20.3 GB of VRAM at Q4_K_M with 8k context, computed from its config.json as checked on 16 October 2026. The smallest card that holds it has 24 GB, and the fastest desktop card that runs it, the RTX 5090, generates about 74 tokens per second.
- Parameters
- 29.3B
- Context
- 131,072 tokens
- Licence
- Apache 2.0
- Released
- August 2026
- Config checked
- 16 October 2026
Open Granite 4.2 30B in the LLM checker
Which card size runs it
20.3 GB against each memory size. A model may use 95% of a card, and only what is left over, or missing, is coloured; tight means above 85%, with little room for a longer conversation.
Other quantisations, context lengths or a Q8 KV cache: the VRAM calculator.
Computed 16 October 2026Which GPU to pick
The smallest card here that loads it at Q4_K_M with 8k of context is the RTX 3090 (24 GB): 20.3 GB needed, an estimated 39 tokens per second. For room to work, meaning 32,768 tokens of context while staying under 85% of memory, step up to the RTX 5090 (32 GB), which needs 26.8 GB for that and generates at about 74 tokens per second.
At Q8_0, which is near-lossless, the card is the RTX A6000 (48 GB), needing 34.9 GB. On Apple silicon, a Mac mini M4 Pro holds it with 32,768 tokens of context at an estimated 11 tokens per second; prompt processing is slower than on Nvidia, which long documents make obvious. Among unified-memory desktops, the Ryzen AI Max+ 395 holds it at 32,768 tokens in 26.8 GB, at an estimated 11 tokens per second.
Speed on each card
| GPU | Fits | Est. tokens/s | Feels like | Memory | Bandwidth |
|---|---|---|---|---|---|
| H100 SXM | Yes | ≤ 138 | faster than you can read | 80 GB | 3,350 GB/s |
| A100 | Yes | 84 | faster than you can read | 80 GB | 2,039 GB/s |
| RTX 5090 | Yes | 74 | faster than you can read | 32 GB | 1,792 GB/s |
| RTX PRO 6000 Blackwell | Yes | 74 | faster than you can read | 96 GB | 1,792 GB/s |
| RTX PRO 5000 Blackwell | Yes | 55 | comfortable for chat | 48 GB | 1,344 GB/s |
| RTX 4090 | Yes | 42 | comfortable for chat | 24 GB | 1,008 GB/s |
| RX 7900 XTX | Yes | 40 | comfortable for chat | 24 GB | 960 GB/s |
| RTX 6000 Ada | Yes | 40 | comfortable for chat | 48 GB | 960 GB/s |
| RTX 3090 | Yes | 39 | comfortable for chat | 24 GB | 936 GB/s |
| RTX PRO 4500 Blackwell | Yes | 37 | comfortable for chat | 32 GB | 896 GB/s |
| L40S | Yes | 36 | comfortable for chat | 48 GB | 864 GB/s |
| RTX A6000 | Yes | 32 | comfortable for chat | 48 GB | 768 GB/s |
| RTX PRO 4000 Blackwell | Yes | 28 | comfortable for chat | 24 GB | 672 GB/s |
| Radeon AI PRO R9700 | Yes | 26 | comfortable for chat | 32 GB | 640 GB/s |
| Arc Pro B70 | Yes | 25 | comfortable for chat | 32 GB | 608 GB/s |
| Arc Pro B60 | Yes | 19 | usable, a little slow | 24 GB | 456 GB/s |
| RTX PRO 4000 Blackwell SFF | Yes | 18 | usable, a little slow | 24 GB | 432 GB/s |
| RTX Spark | Yes | 12 | usable, a little slow | 128 GB (102 for the GPU) | 300 GB/s |
| DGX Spark | Yes | 11 | slow; fine for batch jobs | 128 GB (126 for the GPU) | 273 GB/s |
| Ryzen AI Max+ 395 | Yes | 11 | slow; fine for batch jobs | 128 GB (96 for the GPU) | 256 GB/s |
| RX 7900 XT | No | - | does not fit | 20 GB | 800 GB/s |
| RTX 4060 Ti | No | - | does not fit | 16 GB | 288 GB/s |
| RTX 5060 Ti | No | - | does not fit | 16 GB | 448 GB/s |
| RTX 5080 | No | - | does not fit | 16 GB | 960 GB/s |
| RTX 5070 Ti | No | - | does not fit | 16 GB | 896 GB/s |
| RTX 4080 SUPER | No | - | does not fit | 16 GB | 736 GB/s |
| RX 9070 XT | No | - | does not fit | 16 GB | 640 GB/s |
| RX 9060 XT | No | - | does not fit | 16 GB | 320 GB/s |
| RTX 3060 | No | - | does not fit | 12 GB | 360 GB/s |
| RTX 4070 | No | - | does not fit | 12 GB | 504 GB/s |
| RTX 5070 | No | - | does not fit | 12 GB | 672 GB/s |
| Arc B580 | No | - | does not fit | 12 GB | 456 GB/s |
Single-stream decode, from memory bandwidth at 70% efficiency. Prompt processing and batching not included. These are ceilings, not measurements: Granite 4.2 30B reads only 17 GB of weights per token at Q4, so little that the per-token costs the formula leaves out decide much of the real speed. The figure marked ≤ is where a real runtime falls furthest below the number shown. See the speed estimator for other quantisations and Apple hardware, or every card compared if you are choosing hardware rather than a model.
Computed 16 October 2026Memory by quantisation and context
| Quant | 4,096 ctx | 8,192 ctx | 32,768 ctx |
|---|---|---|---|
| Q8_0 | 33.9 GB · fits 48 GB | 34.9 GB · fits 48 GB | 41.4 GB · fits 48 GB |
| Q5_K_M | 23.2 GB · fits 32 GB | 24.3 GB · fits 32 GB | 30.7 GB · fits 48 GB |
| Q4_K_M | 19.2 GB · fits 24 GB | 20.3 GB · fits 24 GB | 26.8 GB · fits 32 GB |
Weights at this quant: 17 GB at Q4. Every extra 1,000 tokens of context adds about 0.262 GB of KV cache at FP16. Try other settings in the VRAM calculator.
About the model
Granite 4.2 30B has 29.3B parameters, 64 layers and 8 KV heads of dimension 128. The largest of IBM's Granite 4.2 models, August 2026: a dense 30B that thinks before it answers, with a second fine-tuning phase weighted towards agentic data, per the card. Apache 2.0, like the rest of the family.
Text only 131,072-token window Apache 2.0 August 2026
Reasoning. Full thinking by default; a non-thinking mode and a shorter low-effort mode are set per request.
Native context window: 131,072 tokens. Licence: Apache 2.0, a permissive licence. Weights published in August 2026.
Good at
- Agentic tool calling: IBM gave the 30B an extra fine-tuning phase weighted towards agentic data.
- A published list of its training datasets, which most open models do not offer, under Apache 2.0.
- Twelve tested languages, Japanese among them.
Watch out for
- Every layer keeps a full-length cache, so at long context it needs far more memory than Qwen3.8 27B or Gemma 4 31B, which share the work with cheaper layers. On a 24 GB card it fits at Q4 only with a short context.
- Text only: no vision encoder, unlike the Qwen and Gemma models of its size.
- IBM advertises 512K, but that is a long-context extension: the native window is 131,072 tokens, which is what this site counts.
Similar models
The models that need about the same memory as Granite 4.2 30B at Q4_K_M and 8k, which makes them the real alternatives on whatever card you have. The context column is what each extra 1,000 tokens costs, and it is where models of the same size differ most.
| Model | Size | Needs | Per 1k context | Window | Released |
|---|---|---|---|---|---|
| Granite 4.2 30B | 29.3B | 20.3 GB | 0.262 GB | 128k | 2026-08 |
| Gemma 4 31B | 31.3B | 20.9 GB | 0.082 GB | 256k | 2026-03 |
| GLM-4.7 Flash 30B-A3B (MoE) | 31.2B · 3B active | 19.8 GB | 0.054 GB | 198k | 2026-01 |
| Qwen3 30B-A3B (MoE) | 30.5B · 3.3B active | 19.7 GB | 0.098 GB | 40k | 2025-04 |
| Nemotron 3.5 Lightning 30B-A3B (MoE) | 31.6B · 3B active | 19.7 GB | 0.006 GB | 256k | 2026-08 |
See every model compared, which models fit on 8 to 96 GB, or the cheapest hour of every cloud card.
Questions
- Can I run Granite 4.2 30B on a 24 GB card?
- Yes, at Q4_K_M and 8k context it needs about 20.3 GB, which fits a 24 GB card with room.
- How much VRAM does Granite 4.2 30B need at Q8?
- About 34.9 GB at 8k context, or 41.4 GB at 32k. Q8 is near-lossless; use it when it fits.
- What is the context length of Granite 4.2 30B?
- 131,072 tokens natively. Holding all of it at Q4_K_M takes about 52.5 GB, of which 34.4 GB is context.
- Granite 4.2 30B or Qwen3.8 27B?
- Both are dense and Apache 2.0. Qwen3.8 27B reads images and video and pays far less for long context, because only 16 of its 64 layers keep a cache that grows; Granite 4.2 30B is text only and keeps one in all 64. At the same context it needs more memory, and the tables on both pages show how much.
- How fast is Granite 4.2 30B on an RTX 4090?
- At most about 42 tokens per second at Q4, single stream, which is comfortable for chat. That is a ceiling worked out from the card's memory bandwidth, not a measurement.
- Architecture values from config.json in ibm-granite/granite-4.2-30b on Hugging Face, checked 16 October 2026. Every figure on this page is computed from those values; check them before buying hardware for this model.
- Estimates assume a single conversation on a card that is otherwise free. A desktop on the same GPU takes 0.5 to 2 GB.