- Size
- 3.7B parameterscounted from the safetensors header on the Hub
- Licence
- unknown, Apache License 2.0Our reading, check the licence: You can use, change and redistribute the model, including commercially. Keep the licence and any notices, and say which files you changed; you do not have to publish your changes. It includes a patent licence that ends for anyone who sues over patents in the work.
- On the Hub since
- 7 Aug 2026the date the repo was created
- Context length
- 128k tokensfrom the model's config
- Fine-tune
- of ibm-granite/granite-4.1-3b-baseGraniteForCausalLM
- On Hugging Face
- ibm-granite/granite-4.2-3b40k downloads in the last 30 days, 140 likes
Scores
No independent scores for this model in Epoch AI's data. A score is only shown here when Epoch's name for a model matches this repo with certainty; an uncertain match is left out rather than guessed.
Claimed by the maker Claimed
The model card has no claimed scores in a machine-readable form (the card's model-index), so none are shown. The maker may give scores in the card text: read the card.
Other views, which we link to and do not copy: look this model up on Artificial Analysis and the Arena leaderboard.
Run it yourself
Memory needed = weights + context cache + 0.50 GB for the runtime. Context cache at 8k tokens: 0.63 GB, worked out from the model's config (ibm-granite/granite-4.2-3b), at 16-bit precision. Choose a machine above to see a verdict for each format. Method
GGUF for llama.cpp, Ollama, LM Studio and others
ibm-granite/granite-4.2-3b-GGUF (the maker's own conversion), 269k downloads. Sizes are the real files on the Hub, all parts added up.
| Format | Size | Memory needed | Get it |
|---|---|---|---|
| Q3_K_S | 1.6 GBfile size | 2.7 GB1.6 + 0.63 + 0.50 | Commandsllama.cpp llama-server -hf ibm-granite/granite-4.2-3b-GGUF:Q3_K_SOllama ollama run hf.co/ibm-granite/granite-4.2-3b-GGUF:Q3_K_SDownload hf download ibm-granite/granite-4.2-3b-GGUF --include "granite-4.2-3b-Q3_K_S.gguf" --local-dir granite-4.2-3b-GGUF |
| Q3_K_M | 1.7 GBfile size | 2.8 GB1.7 + 0.63 + 0.50 | Commandsllama.cpp llama-server -hf ibm-granite/granite-4.2-3b-GGUF:Q3_K_MOllama ollama run hf.co/ibm-granite/granite-4.2-3b-GGUF:Q3_K_MDownload hf download ibm-granite/granite-4.2-3b-GGUF --include "granite-4.2-3b-Q3_K_M.gguf" --local-dir granite-4.2-3b-GGUF |
| Q3_K_L | 1.8 GBfile size | 3.0 GB1.8 + 0.63 + 0.50 | Commandsllama.cpp llama-server -hf ibm-granite/granite-4.2-3b-GGUF:Q3_K_LOllama ollama run hf.co/ibm-granite/granite-4.2-3b-GGUF:Q3_K_LDownload hf download ibm-granite/granite-4.2-3b-GGUF --include "granite-4.2-3b-Q3_K_L.gguf" --local-dir granite-4.2-3b-GGUF |
| Q4_0 | 2.0 GBfile size | 3.1 GB2.0 + 0.63 + 0.50 | Commandsllama.cpp llama-server -hf ibm-granite/granite-4.2-3b-GGUF:Q4_0Ollama ollama run hf.co/ibm-granite/granite-4.2-3b-GGUF:Q4_0Download hf download ibm-granite/granite-4.2-3b-GGUF --include "granite-4.2-3b-Q4_0.gguf" --local-dir granite-4.2-3b-GGUF |
| Q4_K_S | 2.0 GBfile size | 3.1 GB2.0 + 0.63 + 0.50 | Commandsllama.cpp llama-server -hf ibm-granite/granite-4.2-3b-GGUF:Q4_K_SOllama ollama run hf.co/ibm-granite/granite-4.2-3b-GGUF:Q4_K_SDownload hf download ibm-granite/granite-4.2-3b-GGUF --include "granite-4.2-3b-Q4_K_S.gguf" --local-dir granite-4.2-3b-GGUF |
| Q4_K_M | 2.1 GBfile size | 3.2 GB2.1 + 0.63 + 0.50 | Commandsllama.cpp llama-server -hf ibm-granite/granite-4.2-3b-GGUF:Q4_K_MOllama ollama run hf.co/ibm-granite/granite-4.2-3b-GGUF:Q4_K_MDownload hf download ibm-granite/granite-4.2-3b-GGUF --include "granite-4.2-3b-Q4_K_M.gguf" --local-dir granite-4.2-3b-GGUF |
| Q4_1 | 2.2 GBfile size | 3.3 GB2.2 + 0.63 + 0.50 | Commandsllama.cpp llama-server -hf ibm-granite/granite-4.2-3b-GGUF:Q4_1Ollama ollama run hf.co/ibm-granite/granite-4.2-3b-GGUF:Q4_1Download hf download ibm-granite/granite-4.2-3b-GGUF --include "granite-4.2-3b-Q4_1.gguf" --local-dir granite-4.2-3b-GGUF |
| Q5_0 | 2.4 GBfile size | 3.5 GB2.4 + 0.63 + 0.50 | Commandsllama.cpp llama-server -hf ibm-granite/granite-4.2-3b-GGUF:Q5_0Ollama ollama run hf.co/ibm-granite/granite-4.2-3b-GGUF:Q5_0Download hf download ibm-granite/granite-4.2-3b-GGUF --include "granite-4.2-3b-Q5_0.gguf" --local-dir granite-4.2-3b-GGUF |
| Q5_K_S | 2.4 GBfile size | 3.5 GB2.4 + 0.63 + 0.50 | Commandsllama.cpp llama-server -hf ibm-granite/granite-4.2-3b-GGUF:Q5_K_SOllama ollama run hf.co/ibm-granite/granite-4.2-3b-GGUF:Q5_K_SDownload hf download ibm-granite/granite-4.2-3b-GGUF --include "granite-4.2-3b-Q5_K_S.gguf" --local-dir granite-4.2-3b-GGUF |
| Q5_K_M | 2.4 GBfile size | 3.6 GB2.4 + 0.63 + 0.50 | Commandsllama.cpp llama-server -hf ibm-granite/granite-4.2-3b-GGUF:Q5_K_MOllama ollama run hf.co/ibm-granite/granite-4.2-3b-GGUF:Q5_K_MDownload hf download ibm-granite/granite-4.2-3b-GGUF --include "granite-4.2-3b-Q5_K_M.gguf" --local-dir granite-4.2-3b-GGUF |
| Q5_1 | 2.6 GBfile size | 3.7 GB2.6 + 0.63 + 0.50 | Commandsllama.cpp llama-server -hf ibm-granite/granite-4.2-3b-GGUF:Q5_1Ollama ollama run hf.co/ibm-granite/granite-4.2-3b-GGUF:Q5_1Download hf download ibm-granite/granite-4.2-3b-GGUF --include "granite-4.2-3b-Q5_1.gguf" --local-dir granite-4.2-3b-GGUF |
| Q6_K | 2.8 GBfile size | 3.9 GB2.8 + 0.63 + 0.50 | Commandsllama.cpp llama-server -hf ibm-granite/granite-4.2-3b-GGUF:Q6_KOllama ollama run hf.co/ibm-granite/granite-4.2-3b-GGUF:Q6_KDownload hf download ibm-granite/granite-4.2-3b-GGUF --include "granite-4.2-3b-Q6_K.gguf" --local-dir granite-4.2-3b-GGUF |
| Q8_0 | 3.6 GBfile size | 4.8 GB3.6 + 0.63 + 0.50 | Commandsllama.cpp llama-server -hf ibm-granite/granite-4.2-3b-GGUF:Q8_0Ollama ollama run hf.co/ibm-granite/granite-4.2-3b-GGUF:Q8_0Download hf download ibm-granite/granite-4.2-3b-GGUF --include "granite-4.2-3b-Q8_0.gguf" --local-dir granite-4.2-3b-GGUF |
| BF16 | 6.8 GBfile size | 7.9 GB6.8 + 0.63 + 0.50 | Commandsllama.cpp llama-server -hf ibm-granite/granite-4.2-3b-GGUF:BF16Ollama ollama run hf.co/ibm-granite/granite-4.2-3b-GGUF:BF16Download hf download ibm-granite/granite-4.2-3b-GGUF --include "granite-4.2-3b-bf16.gguf" --local-dir granite-4.2-3b-GGUF |
MLX for Apple silicon
| Format | Size | Memory needed | Repo |
|---|---|---|---|
| MLXibm-granite/granite-4.2-3b-q4-mlx | 1.9 GBfile size | 3.0 GB1.9 + 0.63 + 0.50 | Download page |
| MLXibm-granite/granite-4.2-3b-q8-mlx | 3.6 GBfile size | 4.7 GB3.6 + 0.63 + 0.50 | Download page |
Original weights size from the safetensors header, for vLLM, SGLang, Transformers
| Format | Size | Memory needed | Repo |
|---|---|---|---|
| BF16original weights | 6.8 GBfile size | 7.9 GB6.8 + 0.63 + 0.50 |
hf download ibm-granite/granite-4.2-3b --local-dir granite-4.2-3bOther conversions (7; sizes not read)
- ibm-granite/granite-4.2-3b-fp8 FP8, by the maker, 2.7k downloads
- ibm-granite/granite-4.2-3b-mxfp4 Other, mxfp4, by the maker, 2.5k downloads
- ibm-granite/granite-4.2-3b-nvfp4 NVFP4, by the maker, 2.2k downloads
- litert-community/granite-4.2-3b Other, 632 downloads
- happyme531/Granite-4.2-3B-W4A8-Mixed-QQQ-SQ-RK3588 Other, 101 downloads
- domperidom/granite-4.2-3b-rkllm-rk3576 Other, 66 downloads
- quark75/granite-4.2-3b-W4A16-GPTQ-gs32 GPTQ, w4a16, 24 downloads
The Hub tags 69 quantised repos as conversions of this model; we read the most downloaded and show the maker's own first.
Run it by API
1 provider serves this model through the Hugging Face router. Prices are US dollars per million tokens. Precision is what the provider reports; a lower precision than the original weights can change answers. Live first, then cheapest output.
| Provider | In | Out | Context | Precision | Tokens/s | First token | Uptime 24h | Tools | JSON |
|---|---|---|---|---|---|---|---|---|---|
| deepinfra | $0.03 | $0.12 | 128k | not given | 146 | 0.3 s | ? | yes | no |
Through the router, with a Hugging Face token (the router bills your Hugging Face account). Change the name after the colon to pick another provider.
curl https://router.huggingface.co/v1/chat/completions \
-H "Authorization: Bearer $HF_TOKEN" -H "Content-Type: application/json" \
-d '{"model": "ibm-granite/granite-4.2-3b:deepinfra", "messages": [{"role": "user", "content": "Hello"}]}'Not built yet: one key for every open model, with BlockRun, and our own tested-quality score per provider. Until then, precision and measured speed are the best guide to a provider here.
Planned for this page Not built yet
- Try it in private chat
- A link to Solana AI's confidential chat for the models it serves. This model is not listed as served there.
- Fine-tune it
- Fine-tuning a few base models from this page and publishing the result here.
- Creators and earnings
- Pages for people who publish fine-tunes and adapters of this model, and what they earn.
- Fund and bounties
- Bounties for new open models, paid from a public fund. How the fund will work.
- Contribute compute
- How to give compute to open model work, with Akash.
Model data from the Hugging Face Hub 19 Nov 2026, 13:43 UTC. The repo. Found a mistake on this page? Tell us.