Tencent Hunyuan long-term-memory writer (Oct 15, Apache 2.0), 30B total / ~3B active. Not a general chat model: it slots into memory frameworks like Mem0 / MemOS / Hy-Memory to extract, merge and update long-term memory entries over conversation streams. Trained with OPD + RL; paper arXiv 2610.22196. FP8 official weights and community GGUF conversions are out.
0
Quantized builds
31
Benchmark records
1
Discussions
Quantized builds
Community and official quantized releases; "linked records" counts only benchmarks tied to that exact release.
No quantized builds cataloged yet.
Benchmark data
31 configurations
| Hardware | Framework | Quant | Decode | VRAM used | Samples | Evidence |
|---|---|---|---|---|---|---|
| NVIDIA RTX Spark | llama.cpp | FP16 | 42.8 tok/s | 110.2 GB | 1 | L2 Cross-framework verified |
| NVIDIA RTX Spark | llama.cpp | Q5_K_M | 67.3 tok/s | 60.69 GB | 1 | L2 Cross-framework verified |
| NVIDIA RTX Spark | llama.cpp | Q8_0 | 58.7 tok/s | 74.55 GB | 1 | L2 Cross-framework verified |
| NVIDIA RTX Spark | Ollama | Q5_K_M | 64.3 tok/s | 60.69 GB | 1 | L2 Cross-framework verified |
| NVIDIA RTX Spark | Ollama | Q8_0 | 46.8 tok/s | 74.55 GB | 1 | L2 Cross-framework verified |
| AMD Ryzen AI Max+ 395 | llama.cpp | IQ1_S | 83.7 tok/s | 28.56 GB | 1 | L2 Cross-framework verified |
| AMD Ryzen AI Max+ 395 | vLLM | Q5_K_S | 50.6 tok/s | 58.8 GB | 1 | L2 Cross-framework verified |
| Apple M5 Max | llama.cpp | IQ3_XS | 147.7 tok/s | 41.16 GB | 1 | L2 Cross-framework verified |
| Apple M5 Max | llama.cpp | Q3_K_M | 134.9 tok/s | 44.94 GB | 1 | L2 Cross-framework verified |
| Apple M5 Max | Ollama | IQ2_XS | 139.9 tok/s | 32.34 GB | 1 | L2 Cross-framework verified |
| Apple M5 Max | Ollama | Q8_0 | 63.9 tok/s | 74.55 GB | 1 | L2 Cross-framework verified |
| Apple M5 Ultra | llama.cpp | IQ2_M | 216.9 tok/s | 34.86 GB | 1 | L2 Cross-framework verified |
| Apple M5 Ultra | llama.cpp | Q3_K_S | 189.4 tok/s | 43.05 GB | 1 | L2 Cross-framework verified |
| Apple M5 Ultra | Ollama | Q5_K_S | 139.7 tok/s | 58.8 GB | 1 | L2 Cross-framework verified |
| NVIDIA H100 80GB HBM3 | llama.cpp | Q6_K | 289.8 tok/s | 65.1 GB | 1 | L2 Cross-framework verified |
| NVIDIA H100 80GB HBM3 | vLLM | IQ2_XS | 346.2 tok/s | 32.34 GB | 1 | L2 Cross-framework verified |
| NVIDIA H100 80GB HBM3 | vLLM | Q5_K_M | 299.8 tok/s | 60.69 GB | 1 | L2 Cross-framework verified |
| NVIDIA H20 | llama.cpp | Q6_K | 57.6 tok/s | 65.1 GB | 1 | L2 Cross-framework verified |
| NVIDIA H20 | Strata | IQ1_S | 178.2 tok/s | 28.56 GB | 1 | L2 Cross-framework verified |
| NVIDIA H20 | Strata | IQ3_XS | 139.1 tok/s | 41.16 GB | 1 | L2 Cross-framework verified |
| NVIDIA RTX PRO 6000 Blackwell 96GB | llama.cpp | IQ2_M | 366.3 tok/s | 34.86 GB | 1 | L2 Cross-framework verified |
| NVIDIA RTX PRO 6000 Blackwell 96GB | llama.cpp | IQ3_XS | 355.2 tok/s | 41.16 GB | 1 | L2 Cross-framework verified |
| NVIDIA RTX PRO 6000 Blackwell 96GB | Strata | IQ2_XS | 349.8 tok/s | 32.34 GB | 1 | L2 Cross-framework verified |
| Apple M2 Ultra | llama.cpp | IQ2_XS | 240.8 tok/s | 32.34 GB | 1 | L2 Cross-framework verified |
| Apple M2 Ultra | llama.cpp | Q2_K | 207.8 tok/s | 36.75 GB | 1 | L2 Cross-framework verified |
| Apple M2 Ultra | Ollama | Q2_K | 187.1 tok/s | 36.75 GB | 1 | L2 Cross-framework verified |
| Apple M2 Ultra | Ollama | Q4_K_M | 134.5 tok/s | 53.76 GB | 1 | L2 Cross-framework verified |
| NVIDIA RTX Spark | llama.cpp | IQ4_XS | 93.7 tok/s | 48.72 GB | 1 | L2 Cross-framework verified |
| NVIDIA RTX Spark | llama.cpp | Q4_K_M | 85.8 tok/s | 53.76 GB | 1 | L2 Cross-framework verified |
| NVIDIA RTX Spark | Ollama | IQ4_XS | 76.2 tok/s | 48.72 GB | 1 | L2 Cross-framework verified |
| NVIDIA RTX Spark | Ollama | Q4_K_M | 69.5 tok/s | 53.76 GB | 1 | L2 Cross-framework verified |
31 of these records are not linked to a specific quantized build; the full set is on the search page.