Leanstral — local AI model by Mistral AI
Written by Jakub Rusinowski · Last updated
Mistral's Lean 4 proof-engineering models, on the Mistral Small 4 architecture. Apache-2.0 open weights; the hosted API was retired on 30 September 2026, so the weights are now the only way to use it.
Variants
The smallest Leanstral variant needs about 73 GB of VRAM at Q4_K_M — quantized weights plus framework overhead, before any KV cache.
| Model | VRAM at Q4 | VRAM | Context | Run it |
|---|---|---|---|---|
| Leanstral 1.5 119B-A6B → 119B (6B active) | ~72.6 GB | 262,144 | cloud-hosted tag |
Memory is quantized weights plus overhead at Q4_K_M, from the same engine as the GPU & VRAM checker.
How to run Leanstral locally
Install Ollama, then pull the tag.
No Ollama tag. Mistral's card documents vLLM; its hosted API was retired on 30 September 2026.
Pick a size above for its own VRAM figure, speed estimate and install command.
Licence
Commercial use permitted. No usage restrictions beyond attribution.
Applies to: Leanstral 1.5 119B-A6BRecommended GPU
The cheapest catalogued GPU that runs Leanstral locally (min 73 GB VRAM) is the NVIDIA DGX Spark (128 GB).
Leanstral — frequently asked questions
How much VRAM does Leanstral need?
Leanstral needs about 73 GB VRAM at Q4_K_M quantization for its smallest variant. Variants: Leanstral 1.5 119B-A6B (73 GB, Q4_K_M). On Apple Silicon, unified memory counts toward this requirement.
Can I run Leanstral on an RTX 4090 (24 GB)?
Leanstral's smallest variant needs about 73 GB, which exceeds a single RTX 4090 (24 GB). Use multiple GPUs, a higher-VRAM card, or Apple Silicon with large unified memory.
What quantization should I use for Leanstral?
Q4_K_M is the best balance of quality and VRAM for Leanstral in most cases. Choose Q8_0 for near-lossless quality if you have spare VRAM, or smaller quants (Q3/Q2) only when memory is tight.
How do I run Leanstral with Ollama?
Leanstral has no local Ollama tag — the published tag is cloud-hosted, so running it sends your prompts to a hosted GPU rather than your own machine.