Qwen3.8 Flash-Next
GSQ-RCO IQ3_S GGUF + mmproj + refusal vector · alesha-pro/Qwen3.8-Flash-Next-abliterated-GSQ-RCO-Strata-GGUF
done abliterated
| Publisher | alesha-pro |
|---|---|
| Modality | text+vision LLM |
| Parameters | 125.0B total / 6.0B active 125B MoE with 6B activated, plus 51B n-gram embedding and 4B MTP (model card). 512 experts, 10 routed + 1 shared active. Native context 262,144. Source: model card. |
| Format | GSQ-RCO IQ3_S GGUF + mmproj + refusal vector |
| Size on disk | 78.72 GiB (84,525,749,351 bytes) |
| Estimated memory to run | ~92.8 GB (estimate; see method) |
| Native context | 262,144 tokens (config/card) |
| Engine | Strata engine (setup per card) · recipe |
| License | qwen-community-license-1.0 Qwen Community License 1.0 (inherited from Flash-Next). |
| Archive folder | abliterated |
| Notes | Card reports peak 23.8 GB VRAM on its test rig with Strata offload. |
Abliterated model: refusal behaviour was removed or suppressed by a third party. Use it only for lawful research on systems you own or are authorised to test, keep it off public endpoints, and you own whatever it produces. See responsibility note.
Where it runs
| Machine | Verdict | Budget used |
|---|---|---|
| Bosgame M5 (x1) | Fits ≈75.0 tok/s ceiling | ~108-110 GB GPU-usable on Linux with GTT raised (96 GB cap on Windows). Fit budget used here: 108 GB. |
| Bosgame M5 (x2, RPC) | Fits ≈75.0 tok/s ceiling | ~200-225 GB usable across two nodes. Fit budget used here: 210 GB; 210-225 GB = tight. |
| MacBook Pro (M4 Max) | No | macOS caps GPU wired memory by default (~75% = ~27 GB). Raising iogpu.wired_limit_mb to ~32 GB is possible but leaves little for the OS. |
| Mac mini (M4) | No | ~2/3 of 16 GB usable by the GPU by default. Already busy as build box + Hermes/Ollama. |
| omarchy (AMD desktop) | After 128 GB RAM | 7 GB of the 8 GB VRAM for weights. Offload capacity depends on system RAM, which is unknown today; after the 128 GB upgrade ~110 GB of MoE experts could sit in RAM. |
| Ugreen DXP2800 NAS | No | Storage host. Already runs ollama/open-webui containers; only tiny models (e.g. Whisper on CPU) are realistic. |
| MacBook Neo | Unknown | Specs unknown: machine was offline/unreachable on Nov 14 2026. |
Token/s = theoretical ceiling from memory bandwidth and active weight bytes; not a benchmark. Nothing on this page is a benchmark score.