Model reference · open weights
Qwen3.8-Flash-Next-Mixed-Quant-SSD-PLE is an open-weight language model from Baekpica, listed in the AxForge catalogue. AxForge can bring it up on EU-owned hardware for you on request — with the licence handled where one is required.
About
Qwen3.8-Flash-Next Mixed-Quant SSD-PLE GGUF Two accelerator-resident precision variants are published with one shared BF16 SSD-PLE sidecar. MQ-Q6-SSD-PLE-BF16 is the original Q5/Q6 recipe; MQ-Q5-SSD-PLE-BF16 lowers the Q5K and Q6K targets by one tier, freeing 13.4082 GiB of resident-memory headroom while retaining the optimized Q50 down tail. On a single NVIDIA DGX Spark, the Q5 variant reached 471.13 tok/s mean cold prefill across 6.4K-9.8K-token API prompts. Embedded MTP can be enabled with --mtp-draft 2; in a controlled four-prompt A/B, it increased decode throughput from 23.65 to 28.65 tok/s—up to +21.1%. Native still-image input is also verified end to end across OpenAI Chat Completions, OpenAI Responses, and Anthropic Messages. An 8,243-token multimodal request measured 489.4 tok/s prefill and 22.5 tok/s decode through the model's embedded vision tower. Identical image-bearing prompts can reuse live recurrent checkpoints and disk-KV records through decoded-pixel cache identity; different pixels remain isolated. The Q5 artifact also completed an exact 262,144-token direct prefill at 277.06 tok/s, with all 248,320 final logits finite and the independently recomputed argmax matching the recorded result. An instrumented incremental sweep measured 277.49 tok/s token-weighted prefill and the SSD-PLE latency distribution reported below. The API figures remain guarded, content-sensitive serving checks. The full-window result is a direct no-decode run, not a full-window API-generation or sustained-throughput claim. Support my work I work on making large language models practical on hardware they were never really designed to fit on — through mixed quantization, inference optimization, custom kernels, and serving experiments. While much of the development happens on local hardware, calibration, profiling, and large-scale validation often require expensive on-demand GPUs. Contributions help pay for that compute, storage, and testing infrastructure, so I can keep experimenting and publishing the results openly. This is a memory-hierarchy-aware mixed-precision conversion of Qwen/Qwen3.8-Flash-Next, pinned to revision f5d08274bafd880402bd16f5e3e6c514136ec06c. The design
Summarised from the published model card. Read the full card on the HuggingFace links below.
Specifications
| Maker | Baekpica |
|---|---|
| Type | Language models |
| Variants | 1 |
| Based on | Qwen/Qwen3.8-Flash-Next |
| Released | 2026-08-26 |
| Popularity | 9k downloads / month |
| Likes | 20 |
| Licence | Commercial licence needed |
How it works
Variants
Open weights ship in several sizes and precisions. One page, all the variants — pick the one that fits your GPU. VRAM figures are estimates from model size.
| Variant | Params | Precision | VRAM | Fits 16 GB | Weights |
|---|---|---|---|---|---|
| Qwen3.8-Flash-Next-Mixed-Quant-SSD-PLE-GGUF | — | GGUF | — | — | Weights ↗ |
Using it via the API
Once AxForge deploys qwen3-8-flash-next-mixed-quant-ssd-ple for you, it answers on the OpenAI-compatible API — the same base URL and keys as every other model. (qwen3-8-flash-next-mixed-quant-ssd-ple below is illustrative; you get the exact model name on deployment.)
$ curl -sS https://api.axforge.ai/v1/chat/completions \
-H "Authorization: Bearer $AXFORGE_API_KEY" \
-H "Content-Type: application/json" \
-d '{"model":"qwen3-8-flash-next-mixed-quant-ssd-ple","messages":[{"role":"user","content":"Hello"}]}'
Details
Tags
Licence
The weights are open but its licence needs a commercial agreement for business use. AxForge can arrange that licence and host the model for you — you pay AxForge, we settle with the model’s maker. Ask us for a quote. Read the licence ↗