1× RTX 3090
168tok/sEveryday
1 request, prose prompt, with MTP (per recipe) Source
Intelligence17with thinkingno thinking 13
Local AI · Recipe · 1× RTX 5090
Gemma 4 31B Q6_K with llama.cpp on 1× RTX 5090
Gemma 4 31B Q6_K with llama.cpp on 1× RTX 5090: 53 tok/s according to huggingface.co (dataset as of Nov 3, 2026).
by notwitcheer
Intelligence (original model) · Artificial Analysis19with thinkingno thinking 14
Engine
- Engine
- llama.cpp
- Quantization
- GGUF Q6_K (23.47 GiB)
- Model family
- Gemma 4
- Context
- —
- Parameters
- 31B dense
- Creator
- notwitcheer
- GitHub stars
- 46
- Repo updated
- Nov 1, 2026
Measurements
Every sourced value of this recipe, each with its condition and source. Bars relative to the largest value in the group.
Decode · 1 request
Prefill by context
What you need
- Hardware
- 1 × NVIDIA RTX 5090
- Weights
- Main weightsunsloth/gemma-4-31B-it-GGUF Apache-2.0
- Engine
- llama.cpp
Related recipes
1× RTX 3090
44.9tok/sEveryday
1 request, prose prompt, with DFlash (per recipe) Source
Intelligence19with thinkingno thinking 14
1× DGX Spark
29.8tok/sPeak
1 request, best value reported by source Source
Intelligence17with thinkingno thinking 13
1× RTX 5090
252tok/sEveryday
1 request, prose prompt, no speculative decoding Source
Intelligence18with thinkingno thinking 15
1× RTX 5090
128tok/sEveryday
1 request, prose prompt, with MTP (per recipe) Source
Intelligence21with thinkingno thinking 20
1× RTX 5090
186tok/sMixed
1 request, mixed prompt set, with MTP Source
Intelligence21with thinkingno thinking 20
- Custom kernel required
Local AI in your company?
In a workshop we work out which models and which hardware fit your tasks, and build the first agent on your infrastructure.