AI ToolsJun 28, 2026, 6:35 PM

Ornith-1.0-35B GGUF update: native MTP speculative-decode graft + full serving/TTFT/long-context numbers (llama.cpp, tp=1)

TickrWire Editorial Desk·Jun 28, 2026, 6:35 PM·1 min read AI-assisted, human-reviewed

Reported by the original publisher: Ornith-1.0-35B GGUF update: native MTP speculative-decode graft + full serving/TTFT/long-context numbers (llama.cpp, tp=1). Analysis and context written by TickrWire.

30-second summary

A follow-up update to the Ornith-1.0-35B GGUF model introduces native MTP speculative-decode grafting, achieving 1.3-1.35x single-stream decode speed with identical token distribution to the target model.

TickrWire
Ornith-1.0-35B GGUF update: native MTP speculative-decode graft + full serving/TTFT/long-context numbers (llama.cpp, tp=1)
Full story

Follow-up to my previous Ornith-1.0-35B Q3_K_M post.

I grafted a native MTP draft head onto the IQ4_XS body (head at Q6) for self-speculative decode, single GPU, llama.cpp:

1.3-1.35x single-stream decode (172.6 -> 233.8 tok/s).

Next-token distribution is byte-identical to target-only (KLD 0.0, 32/32).

BF16 KLD 0.073 — slightly better than Q4_K_M.

Issue: not bit-exact to target-only over lon

Sources · 1
Read next
More stories