4 builds of 12b~yuxinlu1 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 256K · proved 128K context proved by yuxinlu1 Q4_K_M
25.8 tok/s peak, by yuxinlu1 Q4_K_M
proved by at least one build: Tool calling · Tool loop finished · Faithful to tool results · Chained tool calls · Structured output · Sustained output · Extended thinking · Writes code · Reasons about code · Synthesises sources · Follows instructions
| build | publisher | precision | engine | whose work | context proved | best tok/s | machines |
|---|---|---|---|---|---|---|---|
| hf.co/yuxinlu1/gemma-4-12B-agentic-fable5-composer2.5-v2-3.5x-tau2-GGUF:Q4_K_M | yuxinlu1 | Q4_K_M | ollama | — | 1,31,072 | 25.8 | 1 |
| hf.co/yuxinlu1/gemma-4-12B-agentic-fable5-composer2.5-v2-3.5x-tau2-GGUF:Q6_K | yuxinlu1 | Q6_K | ollama | — | 1,31,072 | 20.0 | 1 |
| hf.co/yuxinlu1/gemma-4-12B-agentic-fable5-composer2.5-v2-3.5x-tau2-GGUF:Q8_0 | yuxinlu1 | Q8_0 | ollama | — | 1,31,072 | 18.0 | 1 |
| hf.co/yuxinlu1/gemma-4-12B-agentic-fable5-composer2.5-v2-3.5x-tau2-GGUF:Q3_K_M | yuxinlu1 | Q3_K_M | ollama | — | 1,31,072 | 17.0 | 1 |