tiyuvta/Qwen3.8-Flash-Next-NVFP4
Image-Text-to-Text • 117B • Updated • 108
AI & ML interests
Tiyuvta.ai hosts open LLMs with an OpenAI-compatible API tuned for low latency and agent performance. It offers full tool calling, instant API keys, and pay-per-use pricing with no subscriptions or minimums and is targeted at customer support, automation, agent-based products and always-on agents.
Recent Activity
Avifenesh
updated a
model 2 days ago
Avifenesh
updated a
model 4 days ago
Avifenesh
published a
model 4 days ago
Avifenesh
updated a
Space 4 days ago
Avifenesh
published a
Space 5 days ago
Context section correction: 1M demonstrated, chunk-schedule fix, honest boundaries
#4 opened 5 days ago
by 
Avifenesh
Context section correction
#3 opened 5 days ago
by Avifenesh
Context section correction
#2 opened 5 days ago
by Avifenesh
Avifenesh
updated 6
models 5 days ago
tiyuvta/DeepSeek-V4-Flash-0731-NVFP4
Text Generation • 304B • Updated • 123
tiyuvta/memra-bench
32.8k • Updated • 1.53k • 3
tiyuvta/Qwen3.8-27B-DSpark-Agentic
Text Generation • 1B • Updated • 1.08k • 4
tiyuvta/Qwen3.8-27B-DFlash2-memra
Text Generation • 2B • Updated • 1.12k • 2
tiyuvta/Ornith-1.5-9B-NVFP4-MTP-GGUF
Text Generation • 32.8k • Updated • 1.1k • 1
tiyuvta/Qwen3.8-27B-NVFP4-MTP-GGUF
Text Generation • 32.8k • Updated • 11.1k • 14
Post
192
Wrote this up because it still feels backwards. Acceptance dropped, tok/s went up.
Full draft head: 66.7% / 117.1 tok/s. Trimmed: 63.6% / 121.7. Later remeasure +5.1% on a Pro 6000, +6.4% on a 5090 laptop. Full head still accepts more. Loses anyway.
Not my idea. FR-Spec. I just wired it for the native MTP head.
https://huggingface.co/blog/Avifenesh/masked-mtp-drafts
Avifenesh/Qwen3.8-27B-NVFP4-MTP-GGUF
Full draft head: 66.7% / 117.1 tok/s. Trimmed: 63.6% / 121.7. Later remeasure +5.1% on a Pro 6000, +6.4% on a 5090 laptop. Full head still accepts more. Loses anyway.
Not my idea. FR-Spec. I just wired it for the native MTP head.
https://huggingface.co/blog/Avifenesh/masked-mtp-drafts
Avifenesh/Qwen3.8-27B-NVFP4-MTP-GGUF
- 13 replies