IBM · ibm/granite-4-2-30b
| Provider | IBM |
|---|---|
| Family | granite |
| Type | llm-chat |
| Status | active |
| Released | 2026-08-07 |
| Updated | 2026-09-04 |
| Parameters | 29.3B |
| Open weights | yes |
| Relationship | Model |
|---|---|
| Is a finetune of | ibm-granite/granite-4.1-30b-base |
Ordered by predicted speed. Bandwidth sets decode rate; memory decides whether it runs at all.
| Device | Memory | Bandwidth | Best quality | Weights | tok/s | Fastest |
|---|---|---|---|---|---|---|
| Cerebras WSE-3 | 44.0 GB | 21000000.0 GB/s | fp8 | 29.28 GB | ~502104.6 | ~1004209.1 q4 |
| NVIDIA GB200 Grace Blackwell Superchip | 372.0 GB | 16000.0 GB/s | bf16 | 58.55 GB | ~191.3 | ~765.1 q4 |
| AMD Instinct MI355X | 288.0 GB | 8000.0 GB/s | bf16 | 58.55 GB | ~95.6 | ~382.6 q4 |
| NVIDIA B200 | 180.0 GB | 7700.0 GB/s | bf16 | 58.55 GB | ~92.1 | ~368.2 q4 |
| Google TPU7x (Ironwood) | 192.0 GB | 7380.0 GB/s | bf16 | 58.55 GB | ~88.2 | ~352.9 q4 |
| AMD Instinct MI325X | 256.0 GB | 6000.0 GB/s | bf16 | 58.55 GB | ~71.7 | ~286.9 q4 |
| AMD Instinct MI300X | 192.0 GB | 5300.0 GB/s | bf16 | 58.55 GB | ~63.4 | ~253.4 q4 |
| NVIDIA H200 SXM | 141.0 GB | 4800.0 GB/s | bf16 | 58.55 GB | ~57.4 | ~229.5 q4 |
| NVIDIA H100 NVL | 94.0 GB | 3900.0 GB/s | bf16 | 58.55 GB | ~46.6 | ~186.5 q4 |
| NVIDIA H100 SXM | 80.0 GB | 3350.0 GB/s | bf16 | 58.55 GB | ~40.0 | ~160.2 q4 |
| AMD Instinct MI250X | 128.0 GB | 3200.0 GB/s | bf16 | 58.55 GB | ~38.3 | ~153.0 q4 |
| Google TPU v5p | 95.0 GB | 2765.0 GB/s | bf16 | 58.55 GB | ~33.1 | ~132.2 q4 |
| NVIDIA A100 80GB SXM | 80.0 GB | 2039.0 GB/s | bf16 | 58.55 GB | ~24.4 | ~97.5 q4 |
| NVIDIA H100 PCIe | 80.0 GB | 2000.0 GB/s | bf16 | 58.55 GB | ~23.9 | ~95.6 q4 |
| NVIDIA GeForce RTX 5090 | 32.0 GB | 1792.0 GB/s | q6 | 21.96 GB | ~57.1 | ~85.7 q4 |
| Google TPU v6e (Trillium) | 32.0 GB | 1638.0 GB/s | q6 | 21.96 GB | ~52.2 | ~78.3 q4 |
| AMD Instinct MI210 | 64.0 GB | 1600.0 GB/s | fp8 | 29.28 GB | ~38.3 | ~76.5 q4 |
| NVIDIA A100 40GB SXM | 40.0 GB | 1555.0 GB/s | fp8 | 29.28 GB | ~37.2 | ~74.4 q4 |
| Google TPU v4 | 32.0 GB | 1200.0 GB/s | q6 | 21.96 GB | ~38.3 | ~57.4 q4 |
| NVIDIA GeForce RTX 4090 | 24.0 GB | 1008.0 GB/s | q4 | 14.64 GB | ~48.2 | ~48.2 q4 |
| AMD Radeon RX 7900 XTX | 24.0 GB | 960.0 GB/s | q4 | 14.64 GB | ~45.9 | ~45.9 q4 |
| NVIDIA GeForce RTX 3090 | 24.0 GB | 936.0 GB/s | q4 | 14.64 GB | ~44.8 | ~44.8 q4 |
| NVIDIA L40S | 48.0 GB | 864.0 GB/s | fp8 | 29.28 GB | ~20.7 | ~41.3 q4 |
| Apple M3 Ultra | 512.0 GB | 819.0 GB/s | bf16 | 58.55 GB | ~9.8 | ~39.2 q4 |
| AMD Radeon RX 7900 XT | 20.0 GB | 800.0 GB/s | q4 | 14.64 GB | ~38.3 | ~38.3 q4 |
| Apple M2 Ultra | 192.0 GB | 800.0 GB/s | bf16 | 58.55 GB | ~9.6 | ~38.3 q4 |
| Apple M4 Max | 128.0 GB | 546.0 GB/s | bf16 | 58.55 GB | ~6.5 | ~26.1 q4 |
| Apple M1 Max | 64.0 GB | 400.0 GB/s | fp8 | 29.28 GB | ~9.6 | ~19.1 q4 |
| Apple M2 Max | 96.0 GB | 400.0 GB/s | bf16 | 58.55 GB | ~4.8 | ~19.1 q4 |
| Apple M3 Max | 128.0 GB | 400.0 GB/s | bf16 | 58.55 GB | ~4.8 | ~19.1 q4 |
| NVIDIA L4 | 24.0 GB | 300.0 GB/s | q4 | 14.64 GB | ~14.3 | ~14.3 q4 |
| Apple M4 Pro | 64.0 GB | 273.0 GB/s | fp8 | 29.28 GB | ~6.5 | ~13.1 q4 |
| Apple M4 | 32.0 GB | 120.0 GB/s | q6 | 21.96 GB | ~3.8 | ~5.7 q4 |
Taken from the card's availability section.
| Platform | Id on that platform |
|---|---|
| Hugging Face | ibm-granite/granite-4.2-30b |
| Benchmark | Evaluated as | Score | Evidence date | Source kind | |
|---|---|---|---|---|---|
| aa_lcr | Granite 4.2 30B | 49.0% | 2026-09-10 evaluated | independent evaluator | source |
| critpt | Granite 4.2 30B | 0.29% | 2026-09-10 evaluated | independent evaluator | source |
| gdpval_aa | Granite 4.2 30B | 11.05% | 2026-09-10 evaluated | independent evaluator | source |
| gpqa_diamond | Granite 4.2 30B | 64.44% | 2026-09-10 evaluated | independent evaluator | source |
| scicode | Granite 4.2 30B | 37.85% | 2026-09-10 evaluated | independent evaluator | source |
This card reports no benchmark scores yet.
| Benchmark | Catalogue standing | Score |
|---|