An AMD engineer gave Gemini 4 Neon the Inkling 2 Mini model file and let it write, profile and tune GPU kernels on a Strix Halo (8060S) GPU for 4.3 hours and about 490 tool calls. Compared with llama.cpp on ROCm, prefill went from 233 to 1,465 tokens/s and...
VQV Signal
Gemini 4 Neon writes its own AMD GPU kernels, making Inkling 2 Mini run up to 5.9x faster
An AMD engineer gave Gemini 4 Neon the Inkling 2 Mini model file and let it write, profile and tune GPU kernels on a Strix Halo (8060S) GPU for 4.3 hours and about 490 tool calls. Compared with llama.cpp on ROCm, prefill went from 233 to 1,465 tokens/s and...
An AMD engineer gave Gemini 4 Neon the Inkling 2 Mini model file and let it write, profile and tune GPU kernels on a Strix Halo (8060S) GPU for 4.3 hours and about 490 tool calls. Compared with llama.cpp on ROCm, prefill went from 233 to 1,465 tokens/s and decode from 23.8 to 44.8 tokens/s, reaching about 80% of...
Hardware and robotics watchers may want to track whether this becomes a product, benchmark, or deployment signal.
VQV organizes public signals from inspectable sources. It does not independently verify the underlying report.
Signal Strength reflects source quality, relevance, freshness and evidence. Public Interest helps organize discovery; it is not proof of truth.
Public Interest components
VQV surfaced this signal because it is recent, relevant to Consumer AI, connected to Scobleizer News Pulse (X).
No login, cookies, social SDKs, or automatic posting.