Granite Switch 4.1 30B
- Tool use
- Open weights
Prices
0 offerings · prices as listedNo priced offering is listed for this model yet.
Details
- Licence
- apache-2.0
- Family
- granite-switch
Not stated on IBM’s pages: max input, output modalities, reasoning, reasoning efforts, reasoning can be off, max thinking budget, structured output, built-in tools, knowledge cutoff, aliases.
Run locally
Weights: ibm-granite/granite-switch-4.1-30b- Parameters 30B
- Dense
- Native precision BF16
- Layers 66
- KV cache per token 264 KB
- Config positions 128K (config.json max_position_embeddings)
- Needs
- ≈ 17.6 GB (weights 14.8 + KV cache 2.1 + overhead 0.7)
- Usable
- 21.6 GB (90% of its memory)
- Generation speed
- 28–42 tok/s
- Longest context
- 23K (longer doesn’t fit)
- First token
- about 0.97–1.45 s for a 2K prompt
| Precision | Weights | Needs at 8K | On your hardware | Longest context |
|---|---|---|---|---|
| BF16 (native)maker’s file | 55.9 GB | ≈ 60.7 | ✕Doesn’t fit | — |
| Q8_0maker’s file | 31.7 GB | ≈ 35.4 | ○Needs system memory 3–4 tok/s | — |
| FP8 | 27.9 GB | ≈ 31.4 | ○Needs system memory 4–5 tok/s | — |
| Q6_Kmaker’s file | 25.2 GB | ≈ 28.5 | ○Needs system memory 5–7 tok/s | — |
| Q5_K_Mmaker’s file | 22.3 GB | ≈ 25.5 | ○Needs system memory 7–11 tok/s | — |
| Q5_0maker’s file | 21.7 GB | ≈ 24.8 | ○Needs system memory 8–13 tok/s | — |
| Q4_K_Mmaker’s file | 19.6 GB | ≈ 22.6 | ○Needs system memory 15–22 tok/s | — |
| Q4_0maker’s file | 18.4 GB | ≈ 21.3 | ●Runs well 23–34 tok/s | 9K |
| IQ4_XS | 14.8 GB | ≈ 17.6 | ●Runs well 28–42 tok/s | 23K |
| Q3_K_Mmaker’s filequality drops | 16.3 GB | ≈ 19.2 | ●Runs well 26–38 tok/s | 17K |
| IQ3_XXSquality drops | 10.7 GB | ≈ 13.3 | ●Runs well 37–55 tok/s | 40K |
| Q2_Kmaker’s filequality drops | 13.4 GB | ≈ 16.2 | ●Runs well 30–45 tok/s | 29K |
| IQ2_XXSquality drops | 7.2 GB | ≈ 9.8 | ●Runs well 51–76 tok/s | 54K |
Click a row to choose a precision. “Maker’s file” sizes are the maker’s own files; others are parameters × bits per weight.
Compared with other hardware
Each at the highest precision it runs. Click one for details.
Smallest that runs it (Q4_K_M, 8K): GPU GeForce RTX 5090; unified memory Apple M5 Max (32-core GPU) · 36 GB.
Run commands
The maker publishes no files at this precision. Choose its native precision or one of its own builds, or quantise it yourself with llama.cpp.
Memory to fine-tune
Batch 1, 2,048-token sequences, gradient checkpointing; LoRA and QLoRA train rank-16 adapters on the attention projections.
Estimates; facts from the maker’s Hugging Face repositories, as of 2026-11-06. How it’s estimated
Cited scores
From third-party leaderboards, not measured by usThis model isn’t on any leaderboard we cite yet.
Changes
All changes →- New language model
Sources
- ibm-granite/granite-switch-4.1-30b model card (opens in a new tab) huggingface.co · checked Nov 2, 2026
- ibm-granite/granite-switch-4.1-30b config.json (opens in a new tab) huggingface.co · checked Nov 2, 2026
- ibm-granite/granite-switch-4.1-30b model card (opens in a new tab) huggingface.co · checked Nov 2, 2026
- ibm-granite/granite-switch-4.1-30b files (opens in a new tab) huggingface.co · checked Nov 2, 2026
- ibm-granite/granite-switch-4.1-30b-GGUF files (opens in a new tab) huggingface.co · checked Nov 2, 2026
Some values on this page are still waiting to be checked against official pages: a filled dot is verified, a hollow one is not yet (with or without a source on file). How we verify data
IBM · United States · ibm.com
Related models
Snapshots, the same family and other models from this provider