Mistral AI
Mistral 7B 7B
7,2B
Parameters
4,1 Go
Q4_K_M file
≥ 5,6 Go
Recommended VRAM
32 768
Max context
Speed by graphics card
Simulate in the “Configurator” →Standard llama-bench run (prompt 512 / generation 128 tokens), model fully in VRAM. Click a card for details. Electricity cost at the default price (0,25 €/kWh).
| Card · Q4_K_M | VRAM | Generation | Prompt | VRAM used | Power | € elec. / M tok | Price |
|---|---|---|---|---|---|---|---|
| RTX 4070 Ti | 12 Go | 98,1 tok/s | 6 106 | 4,8 Go | 150 W | 0,11 €/M tok | — |
Not measured on these cards yet — click for an estimate:
Technical sheet
- Published
- 2024-05-22
- Architecture
- 32 layers · 8 KV heads · dimension 128
- Attention
- Standard: every layer keeps the whole context in memory.
- Context cache (f16)
- ≈ 125 MB per 1,000 tokens · 4,0 Go for 32,768 tokens
- GGUF file
- bartowski/Mistral-7B-Instruct-v0.3-GGUF · Mistral-7B-Instruct-v0.3-Q4_K_M.gguf Download (4,1 Go)
- Official repo
- mistralai/Mistral-7B-Instruct-v0.3
Public Hugging Face data (official config.json, GGUF repo). The context cache only counts full-attention layers.