5 builds of Granite 4.0 1B measured on real machines — different packagers, precisions and engines, each a different file with its own numbers. The best each engine proved is first; every build has its own page with the machines and signatures behind it.
trained 128K · proved 4K context proved by ibm-granite Q4_0
11.9 tok/s peak, by ibm-granite Q4_0
proved by at least one build: Tool calling · Tool loop finished · Structured output · Sustained output · Writes code · Follows instructions
| build | publisher | precision | engine | whose work | context proved | best tok/s | machines |
|---|---|---|---|---|---|---|---|
| hf.co/ibm-granite/granite-4.0-1b-GGUF:Q4_0 | ibm-granite | Q4_0 | ollama | vendor | 4,096 | 11.9 | 1 |
| hf.co/ibm-granite/granite-4.0-1b-GGUF:Q4_K_M | ibm-granite | Q4_K_M | ollama | vendor | 4,096 | 11.1 | 1 |
| hf.co/ibm-granite/granite-4.0-1b-GGUF:Q3_K_S | ibm-granite | Q3_K_S | ollama | vendor | 4,096 | 8.5 | 1 |
| hf.co/ibm-granite/granite-4.0-1b-GGUF:Q3_K_L | ibm-granite | Q3_K_L | ollama | vendor | 4,096 | 7.2 | 1 |
| hf.co/ibm-granite/granite-4.0-1b-GGUF:Q2_K | ibm-granite | Q2_K | ollama | vendor | — | 8.2 | 1 |
Fine-tunes, merges and abliterations of this member, each its own model with its own page. None of their numbers are counted above.