Qwen3.8-Flash-Next
Released2026-08-26Size180B · 6B activeArchitecturehybrid linear attnLicenseqwen-community-1.0AA rank#8 of 75
Files and memory at 8K context
| Quant | File | Files | + KV + runtime | Smallest fit | Source |
|---|---|---|---|---|---|
| Q8_0 unsloth/Qwen3.8-Flash-Next-GGUF | 175 GiB | 6 | unknown | 192 GB GPU (cache unknown) | HF files |
| BF16 unsloth/Qwen3.8-Flash-Next-GGUF | 330 GiB | 8 | unknown | over 192 GB GPU | HF files |
| MLX 4-bit mixed precision mlx-community/Qwen3.8-Flash-Next-4bit | 104 GiB | 22 | unknown | 192 GB Mac (cache unknown) | HF files |
No verified file: Q4_K_M, Q5_K_M, FP16, MLX 8-bit. Sizes are exact file bytes (text weights; projectors listed separately). “+” = recurrent state or other terms unknown. Smallest fit adds 0.5 GiB spare; Macs use 75% of memory.
Quality
Artificial Analysis Intelligence Index
39.8
Variant: Qwen3.8-Flash-Next · observed 2026-10-17 · source
Rank #8 of 75 researched 2026 models.
Arena text rating
Unknown
Scores describe the evaluated hosted variant, not each quantized file. AA and Arena are never averaged.
Reported speeds 7
| Hardware | Quant | Backend | Prompt t/s | Gen t/s | Date | Source |
|---|---|---|---|---|---|---|
| Apple M3 Ultra (80-core GPU), 512 GB unified memory | 4bit (Qwen3.8-Flash-Next-oQ4e-mtp) | MLX (oMLX server) | 1,409 | 95.6 | 2026-10-16 | omlx.ai |
| Apple M5 Ultra (80-core GPU), 256 GB unified memory | 8bit (Qwen3.8-Flash-Next-oQ8e-mtp) | MLX (oMLX server) | 474 | 62.6 | 2026-10-16 | omlx.ai |
| Apple M5 Ultra (80-core GPU), 256 GB unified memory | 8bit (Qwen3.8-Flash-Next-oQ8e-mtp) | MLX (oMLX server) | 1,316 | 50 | 2026-10-16 | omlx.ai |
| AMD Ryzen AI MAX+ 395 (Strix Halo), 128GB LPDDR5X, Radeon 8060S iGPU, ROCm 10.0 HIP with hipCUB fix | AtomicChat AD-4.27bpw (92.9 GB) - author tested this and Unsloth UD-Q4_K_XL; quant of the depth table inferred as the production AD-4.27bpw | llama.cpp ROCm | — | 20.6 | 2026-08-29 | sleepingrobots.com |
| AMD Ryzen AI MAX+ 395 (Strix Halo), 128GB LPDDR5X, Radeon 8060S iGPU, ROCm 10.0 HIP with hipCUB fix | AtomicChat AD-4.27bpw (92.9 GB) - author tested this and Unsloth UD-Q4_K_XL; quant of the depth table inferred as the production AD-4.27bpw | llama.cpp ROCm | — | 20.1 | 2026-08-29 | sleepingrobots.com |
| AMD Ryzen AI MAX+ 395 (Strix Halo), 128GB LPDDR5X, Radeon 8060S iGPU, ROCm 10.0 HIP with hipCUB fix | AtomicChat AD-4.27bpw (92.9 GB) - author tested this and Unsloth UD-Q4_K_XL; quant of the depth table inferred as the production AD-4.27bpw | llama.cpp ROCm | — | 17.5 | 2026-08-29 | sleepingrobots.com |
| AMD Ryzen AI MAX+ 395 (Strix Halo), 128GB LPDDR5X, Radeon 8060S iGPU, ROCm 10.0 HIP with hipCUB fix | AtomicChat AD-4.27bpw (92.9 GB) - author tested this and Unsloth UD-Q4_K_XL; quant of the depth table inferred as the production AD-4.27bpw | llama.cpp ROCm | 284 | 15.3 | 2026-08-29 | sleepingrobots.com |
Reported by the linked sources, not reproduced by LocalFit. All rows in the speed database