ThinkingCap-Qwen3.8-27B
30 records across 7 hardware.
| hardware | model | quant | backend | tok/s | pp tok/s | W | ttft s | ctx/tg | date | class | flags | source |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| M5 Ultra | ThinkingCap-Qwen3.8-27B | 8bit | oMLX | 101.7 | 1664.0 | – | – | 32768 | 2026-10-12 | community | – | source |
| RX 7900 XTX | ThinkingCap-Qwen3.8-27B | Q4_K_S | llama.cpp | 92.4 | – | – | – | 65536 | 2026-09-29 | community | – | source |
| M5 Ultra | ThinkingCap-Qwen3.8-27B | 8bit | oMLX | 91.1 | 970.9 | – | – | 32768 | 2026-10-11 | community | contradiction | source |
| RX 7900 XTX | ThinkingCap-Qwen3.8-27B | Q4_K_S | llama.cpp | 81.5 | – | – | – | 65536 | 2026-09-29 | community | – | source |
| RX 7900 XTX | ThinkingCap-Qwen3.8-27B | Q4_K_S | llama.cpp | 78.9 | – | – | – | 65536 | 2026-09-29 | community | – | source |
| M3 Ultra | ThinkingCap-Qwen3.8-27B | oQ6e | oMLX | 45.9 | 299.9 | – | – | 32768 | 2026-10-10 | community | – | source |
| RX 7900 XTX | ThinkingCap-Qwen3.8-27B | Q4_K_S | llama.cpp | 45.6 | – | – | – | 65536 | 2026-09-29 | community | – | source |
| M3 Ultra | ThinkingCap-Qwen3.8-27B | oQ6e | oMLX | 44.9 | 378.8 | – | – | 32768 | 2026-10-11 | community | – | source |
| M3 Ultra | ThinkingCap-Qwen3.8-27B | oQ5e | oMLX | 38.3 | 337.2 | – | – | 4096 | 2026-10-13 | community | – | source |
| M4 Max | ThinkingCap-Qwen3.8-27B | oQ8e | oMLX | 37.7 | 249.5 | – | – | 1024 | 2026-10-08 | community | – | source |
| RX 7900 XTX | ThinkingCap-Qwen3.8-27B | Q4_K_S | llama.cpp | 34.2 | – | – | – | 65536 | 2026-09-29 | community | – | source |
| M4 Max | ThinkingCap-Qwen3.8-27B | oQ8e | oMLX | 31.9 | 254.9 | – | – | 4096 | 2026-10-08 | community | – | source |
| M5 Ultra | ThinkingCap-Qwen3.8-27B | 8bit | oMLX | 30.6 | 1717.0 | – | – | 32768 | 2026-10-11 | community | contradiction | source |
| RX 7900 XTX | ThinkingCap-Qwen3.8-27B | Q4_K_S | llama.cpp | 26.7 | – | – | – | 65536 | 2026-09-29 | community | contradiction | source |
| M3 Ultra | ThinkingCap-Qwen3.8-27B | oQ5e | oMLX | 21.8 | 321.9 | – | – | 16384 | 2026-10-13 | community | – | source |
| M4 Pro | ThinkingCap-Qwen3.8-27B | oQ5e | oMLX | 21.4 | 95.9 | – | – | 1024 | 2026-10-16 | community | – | source |
| M4 Pro | ThinkingCap-Qwen3.8-27B | oQ5e | oMLX | 19.1 | 98.8 | – | – | 1024 | 2026-10-16 | community | contradiction | source |
| M4 Pro | ThinkingCap-Qwen3.8-27B | oQ5e | oMLX | 19.0 | 99.7 | – | – | 4096 | 2026-10-16 | community | – | source |
| M5 Max | ThinkingCap-Qwen3.8-27B | mxfp8 | oMLX | 18.6 | 822.0 | – | – | 1024 | 2026-10-17 | community | – | source |
| M5 Max | ThinkingCap-Qwen3.8-27B | mxfp8 | oMLX | 18.5 | 945.9 | – | – | 4096 | 2026-10-17 | community | – | source |
| M4 Pro | ThinkingCap-Qwen3.8-27B | oQ5e | oMLX | 18.4 | 96.9 | – | – | 4096 | 2026-10-16 | community | – | source |
| M5 Max | ThinkingCap-Qwen3.8-27B | mxfp8 | oMLX | 18.2 | 946.6 | – | – | 8192 | 2026-10-17 | community | – | source |
| M5 Max | ThinkingCap-Qwen3.8-27B | mxfp8 | oMLX | 17.8 | 924.2 | – | – | 16384 | 2026-10-17 | community | – | source |
| M5 Max | ThinkingCap-Qwen3.8-27B | mxfp8 | oMLX | 17.2 | 867.5 | – | – | 32768 | 2026-10-17 | community | – | source |
| M5 Max | ThinkingCap-Qwen3.8-27B | mxfp8 | oMLX | 15.9 | 752.2 | – | – | 65536 | 2026-10-17 | community | – | source |
| RX 7900 XTX | ThinkingCap-Qwen3.8-27B | Q4_K_S | llama.cpp | 15.5 | – | – | – | 65536 | 2026-09-29 | community | – | source |
| M4 Pro | ThinkingCap-Qwen3.8-27B | oQ5e | oMLX | 14.3 | 65.2 | – | – | 1024 | 2026-10-16 | community | contradiction | source |
| M4 Pro | ThinkingCap-Qwen3.8-27B | oQ5e | oMLX | 12.0 | 77.3 | – | – | 4096 | 2026-10-16 | community | contradiction | source |
| M1 Max | ThinkingCap-Qwen3.8-27B | oQ6e | oMLX | 10.8 | 67.7 | – | – | 8192 | 2026-10-09 | community | – | source |
| RX 7900 XTX | ThinkingCap-Qwen3.8-27B | Q4_K_S | llama.cpp | 10.6 | – | – | – | 65536 | 2026-09-29 | community | – | source |